initial commit

.gitignore

@@ -0,0 +1,7 @@
+*.pyc
+*.egg-info
+dist
+datasets
+pytorch_env
+models
+build

README.md

@@ -0,0 +1,94 @@
+# RAFT
+This repository contains the source code for our paper:
+
+[RAFT: Recurrent All Pairs Field Transforms for Optical Flow](https://arxiv.org/pdf/2003.12039.pdf)<br/>
+Zachary Teed and Jia Deng<br/>
+
+## Requirements
+Our code was tested using PyTorch 1.3.1 and Python 3. The following additional packages need to be installed
+
+  ```Shell
+  pip install Pillow
+  pip install scipy
+  pip install opencv-python
+  ```
+
+## Demos
+Pretrained models can be downloaded by running
+```Shell
+./scripts/download_models.sh
+```
+
+You can run the demos using one of the available models.
+
+```Shell
+python demo.py --model=models/chairs+things.pth
+```
+
+or using the small (1M parameter) model
+
+```Shell
+python demo.py --model=models/small.pth --small
+```
+
+Running the demos will display the two images and a vizualization of the optical flow estimate. After the images display, press any key to continue.
+
+## Training
+To train RAFT, you will need to download the required datasets. The first stage of training requires the [FlyingChairs](https://lmb.informatik.uni-freiburg.de/resources/datasets/FlyingChairs.en.html#flyingchairs) and [FlyingThings3D](https://lmb.informatik.uni-freiburg.de/resources/datasets/SceneFlowDatasets.en.html) datasets. Finetuning and evaluation require the [Sintel](http://sintel.is.tue.mpg.de/) and [KITTI](http://www.cvlibs.net/datasets/kitti/eval_scene_flow.php?benchmark=flow) datasets. We organize the directory structure as follows. By default `datasets.py` will search for the datasets in these locations
+
+```Shell
+├── datasets
+│   ├── Sintel
+|   |   ├── test
+|   |   ├── training
+│   ├── KITTI
+|   |   ├── testing
+|   |   ├── training
+|   |   ├── devkit
+│   ├── FlyingChairs_release
+|   |   ├── data
+│   ├── FlyingThings3D
+|   |   ├── frames_cleanpass
+|   |   ├── frames_finalpass
+|   |   ├── optical_flow
+```
+
+We used the following training schedule in our paper (note: we use 2 GPUs for training)
+
+```Shell
+python train.py --name=chairs --image_size 368 496 --dataset=chairs --num_steps=100000 --lr=0.0002 --batch_size=6
+```
+
+Next, finetune on the FlyingThings dataset
+
+```Shell
+python train.py --name=things --image_size 368 768 --dataset=things --num_steps=60000 --lr=0.00005 --batch_size=3 --restore_ckpt=checkpoints/chairs.pth
+```
+
+You can perform dataset specific finetuning
+
+### Sintel
+
+```Shell
+python train.py --name=sintel_ft --image_size 368 768 --dataset=sintel --num_steps=60000 --lr=0.00005 --batch_size=4 --restore_ckpt=checkpoints/things.pth
+```
+
+### KITTI
+
+```Shell
+python train.py --name=kitti_ft --image_size 288 896 --dataset=kitti --num_steps=40000 --lr=0.0001 --batch_size=4 --restore_ckpt=checkpoints/things.pth
+```
+
+
+## Evaluation
+You can evaluate a model on Sintel and KITTI by running
+
+```Shell
+python evaluate.py --model=checkpoints/chairs+things.pth
+```
+
+or the small model by including the `small` flag
+
+```Shell
+python evaluate.py --model=checkpoints/small.pth --small
+```

core/datasets.py

@@ -0,0 +1,312 @@
+# Data loading based on https://github.com/NVIDIA/flownet2-pytorch
+
+import numpy as np
+import torch
+import torch.utils.data as data
+import torch.nn.functional as F
+
+import os
+import cv2
+import math
+import random
+from glob import glob
+import os.path as osp
+
+from utils import frame_utils
+from utils.augmentor import FlowAugmentor, FlowAugmentorKITTI
+
+
+class CombinedDataset(data.Dataset):
+    def __init__(self, datasets):
+        self.datasets = datasets
+
+    def __len__(self):
+        length = 0 
+        for i in range(len(self.datasets)):
+            length += len(self.datsaets[i])
+        return length
+
+    def __getitem__(self, index):
+        i = 0
+        for j in range(len(self.datasets)):
+            if i + len(self.datasets[j]) >= index:
+                yield self.datasets[j][index-i]
+                break
+            i += len(self.datasets[j])
+
+    def __add__(self, other):
+        self.datasets.append(other)
+        return self
+
+class FlowDataset(data.Dataset):
+    def __init__(self, args, image_size=None, do_augument=False):
+        self.image_size = image_size
+        self.do_augument = do_augument
+
+        if self.do_augument:
+            self.augumentor = FlowAugmentor(self.image_size)
+
+        self.flow_list = []
+        self.image_list = []
+
+        self.init_seed = False
+
+    def __getitem__(self, index):
+
+        if not self.init_seed:
+            worker_info = torch.utils.data.get_worker_info()
+            if worker_info is not None:
+                torch.manual_seed(worker_info.id)
+                np.random.seed(worker_info.id)
+                random.seed(worker_info.id)
+                self.init_seed = True
+
+        index = index % len(self.image_list)
+        flow = frame_utils.read_gen(self.flow_list[index])
+        img1 = frame_utils.read_gen(self.image_list[index][0])
+        img2 = frame_utils.read_gen(self.image_list[index][1])
+
+        img1 = np.array(img1).astype(np.uint8)[..., :3]
+        img2 = np.array(img2).astype(np.uint8)[..., :3]
+        flow = np.array(flow).astype(np.float32)
+
+        if self.do_augument:
+            img1, img2, flow = self.augumentor(img1, img2, flow)
+
+        img1 = torch.from_numpy(img1).permute(2, 0, 1).float()
+        img2 = torch.from_numpy(img2).permute(2, 0, 1).float()
+        flow = torch.from_numpy(flow).permute(2, 0, 1).float()
+        valid = torch.ones_like(flow[0])
+
+        return img1, img2, flow, valid
+
+    def __len__(self):
+        return len(self.image_list)
+
+    def __add(self, other):
+        return CombinedDataset([self, other])
+
+
+class MpiSintelTest(FlowDataset):
+    def __init__(self, args, root='datasets/Sintel/test', dstype='clean'):
+        super(MpiSintelTest, self).__init__(args, image_size=None, do_augument=False)
+
+        self.root = root
+        self.dstype = dstype
+
+        image_dir = osp.join(self.root, dstype)
+        all_sequences = os.listdir(image_dir)
+
+        self.image_list = []
+        for sequence in all_sequences:
+            frames = sorted(glob(osp.join(image_dir, sequence, '*.png')))
+            for i in range(len(frames)-1):
+                self.image_list += [[frames[i], frames[i+1], sequence, i]]
+
+    def __getitem__(self, index):
+        img1 = frame_utils.read_gen(self.image_list[index][0])
+        img2 = frame_utils.read_gen(self.image_list[index][1])
+        sequence = self.image_list[index][2]
+        frame = self.image_list[index][3]
+
+        img1 = np.array(img1).astype(np.uint8)[..., :3]
+        img2 = np.array(img2).astype(np.uint8)[..., :3]
+        img1 = torch.from_numpy(img1).permute(2, 0, 1).float()
+        img2 = torch.from_numpy(img2).permute(2, 0, 1).float()
+        return img1, img2, sequence, frame
+
+
+class MpiSintel(FlowDataset):
+    def __init__(self, args, image_size=None, do_augument=True, root='datasets/Sintel/training', dstype='clean'):
+        super(MpiSintel, self).__init__(args, image_size, do_augument)
+        if do_augument:
+            self.augumentor.min_scale = -0.2
+            self.augumentor.max_scale = 0.7
+
+        self.root = root
+        self.dstype = dstype
+
+        flow_root = osp.join(root, 'flow')
+        image_root = osp.join(root, dstype)
+
+        file_list = sorted(glob(osp.join(flow_root, '*/*.flo')))
+        for flo in file_list:
+            fbase = flo[len(flow_root)+1:]
+            fprefix = fbase[:-8]
+            fnum = int(fbase[-8:-4])
+
+            img1 = osp.join(image_root, fprefix + "%04d"%(fnum+0) + '.png')
+            img2 = osp.join(image_root, fprefix + "%04d"%(fnum+1) + '.png')
+
+            if not osp.isfile(img1) or not osp.isfile(img2) or not osp.isfile(flo):
+                continue
+
+            self.image_list.append((img1, img2))
+            self.flow_list.append(flo)
+
+
+class FlyingChairs(FlowDataset):
+    def __init__(self, args, image_size=None, do_augument=True, root='datasets/FlyingChairs_release/data'):
+        super(FlyingChairs, self).__init__(args, image_size, do_augument)
+        self.root = root
+        self.augumentor.min_scale = -0.2
+        self.augumentor.max_scale = 1.0
+
+        images = sorted(glob(osp.join(root, '*.ppm')))
+        self.flow_list = sorted(glob(osp.join(root, '*.flo')))
+        assert (len(images)//2 == len(self.flow_list))
+
+        self.image_list = []
+        for i in range(len(self.flow_list)):
+            im1 = images[2*i]
+            im2 = images[2*i + 1]
+            self.image_list.append([im1, im2])
+
+
+class SceneFlow(FlowDataset):
+    def __init__(self, args, image_size, do_augument=True, root='datasets',
+            dstype='frames_cleanpass', use_flyingthings=True, use_monkaa=False, use_driving=False):
+        
+        super(SceneFlow, self).__init__(args, image_size, do_augument)
+        self.root = root
+        self.dstype = dstype
+
+        self.augumentor.min_scale = -0.2
+        self.augumentor.max_scale = 0.8
+
+        if use_flyingthings:
+            self.add_flyingthings()
+        
+        if use_monkaa:
+            self.add_monkaa()
+
+        if use_driving:
+            self.add_driving()
+
+    def add_flyingthings(self):
+        root = osp.join(self.root, 'FlyingThings3D')
+
+        for cam in ['left']:
+            for direction in ['into_future', 'into_past']:
+                image_dirs = sorted(glob(osp.join(root, self.dstype, 'TRAIN/*/*')))
+                image_dirs = sorted([osp.join(f, cam) for f in image_dirs])
+
+                flow_dirs = sorted(glob(osp.join(root, 'optical_flow/TRAIN/*/*')))
+                flow_dirs = sorted([osp.join(f, direction, cam) for f in flow_dirs])
+
+                for idir, fdir in zip(image_dirs, flow_dirs):
+                    images = sorted(glob(osp.join(idir, '*.png')) )
+                    flows = sorted(glob(osp.join(fdir, '*.pfm')) )
+                    for i in range(len(flows)-1):
+                        if direction == 'into_future':
+                            self.image_list += [[images[i], images[i+1]]]
+                            self.flow_list += [flows[i]]
+                        elif direction == 'into_past':
+                            self.image_list += [[images[i+1], images[i]]]
+                            self.flow_list += [flows[i+1]]
+      
+    def add_monkaa(self):
+        pass # we don't use monkaa
+
+    def add_driving(self):
+        pass # we don't use driving
+
+
+class KITTI(FlowDataset):
+    def __init__(self, args, image_size=None, do_augument=True, is_test=False, is_val=False, do_pad=False, split=True, root='datasets/KITTI'):
+        super(KITTI, self).__init__(args, image_size, do_augument)
+        self.root = root
+        self.is_test = is_test
+        self.is_val = is_val
+        self.do_pad = do_pad
+
+        if self.do_augument:
+            self.augumentor = FlowAugumentorKITTI(self.image_size, args.eraser_aug, min_scale=-0.2, max_scale=0.5)
+
+        if self.is_test:
+            images1 = sorted(glob(os.path.join(root, 'testing', 'image_2/*_10.png')))
+            images2 = sorted(glob(os.path.join(root, 'testing', 'image_2/*_11.png')))
+            for i in range(len(images1)):
+                self.image_list += [[images1[i], images2[i]]]
+
+        else:
+            flows = sorted(glob(os.path.join(root, 'training', 'flow_occ/*_10.png')))
+            images1 = sorted(glob(os.path.join(root, 'training', 'image_2/*_10.png')))
+            images2 = sorted(glob(os.path.join(root, 'training', 'image_2/*_11.png')))
+
+            for i in range(len(flows)):
+                self.flow_list += [flows[i]]
+                self.image_list += [[images1[i], images2[i]]]
+
+
+    def __getitem__(self, index):
+
+        if self.is_test:
+            frame_id = self.image_list[index][0]
+            frame_id = frame_id.split('/')[-1]
+
+            img1 = frame_utils.read_gen(self.image_list[index][0])
+            img2 = frame_utils.read_gen(self.image_list[index][1])
+
+            img1 = np.array(img1).astype(np.uint8)[..., :3]
+            img2 = np.array(img2).astype(np.uint8)[..., :3]
+
+            img1 = torch.from_numpy(img1).permute(2, 0, 1).float()
+            img2 = torch.from_numpy(img2).permute(2, 0, 1).float()
+            return img1, img2, frame_id
+
+
+        else:
+            if not self.init_seed:
+                worker_info = torch.utils.data.get_worker_info()
+                if worker_info is not None:
+                    np.random.seed(worker_info.id)
+                    random.seed(worker_info.id)
+                    self.init_seed = True
+
+            index = index % len(self.image_list)
+            frame_id = self.image_list[index][0]
+            frame_id = frame_id.split('/')[-1]
+
+            img1 = frame_utils.read_gen(self.image_list[index][0])
+            img2 = frame_utils.read_gen(self.image_list[index][1])
+            flow, valid = frame_utils.readFlowKITTI(self.flow_list[index])
+
+            img1 = np.array(img1).astype(np.uint8)[..., :3]
+            img2 = np.array(img2).astype(np.uint8)[..., :3]
+
+            if self.do_augument:
+                img1, img2, flow, valid = self.augumentor(img1, img2, flow, valid)
+
+            img1 = torch.from_numpy(img1).permute(2, 0, 1).float()
+            img2 = torch.from_numpy(img2).permute(2, 0, 1).float()
+            flow = torch.from_numpy(flow).permute(2, 0, 1).float()
+            valid = torch.from_numpy(valid).float()
+
+            if self.do_pad:
+                ht, wd = img1.shape[1:]
+                pad_ht = (((ht // 8) + 1) * 8 - ht) % 8
+                pad_wd = (((wd // 8) + 1) * 8 - wd) % 8
+                pad_ht1 = [0, pad_ht]
+                pad_wd1 = [pad_wd//2, pad_wd - pad_wd//2]
+                pad = pad_wd1 + pad_ht1
+
+                img1 = img1.view(1, 3, ht, wd)
+                img2 = img2.view(1, 3, ht, wd)
+                flow = flow.view(1, 2, ht, wd)
+                valid = valid.view(1, 1, ht, wd)
+
+                img1 = torch.nn.functional.pad(img1, pad, mode='replicate')
+                img2 = torch.nn.functional.pad(img2, pad, mode='replicate')
+                flow = torch.nn.functional.pad(flow, pad, mode='constant', value=0)
+                valid = torch.nn.functional.pad(valid, pad, mode='replicate', value=0)
+
+                img1 = img1.view(3, ht+pad_ht, wd+pad_wd)
+                img2 = img2.view(3, ht+pad_ht, wd+pad_wd)
+                flow = flow.view(2, ht+pad_ht, wd+pad_wd)
+                valid = valid.view(ht+pad_ht, wd+pad_wd)
+
+            if self.is_test:
+                return img1, img2, flow, valid, frame_id
+
+            return img1, img2, flow, valid

core/modules/corr.py

@@ -0,0 +1,53 @@
+import torch
+import torch.nn.functional as F
+from utils.utils import bilinear_sampler, coords_grid
+
+class CorrBlock:
+    def __init__(self, fmap1, fmap2, num_levels=4, radius=4):
+        self.num_levels = num_levels
+        self.radius = radius
+        self.corr_pyramid = []
+
+        # all pairs correlation
+        corr = CorrBlock.corr(fmap1, fmap2)
+
+        batch, h1, w1, dim, h2, w2 = corr.shape
+        corr = corr.view(batch*h1*w1, dim, h2, w2)
+        
+        self.corr_pyramid.append(corr)
+        for i in range(self.num_levels):
+            corr = F.avg_pool2d(corr, 2, stride=2)
+            self.corr_pyramid.append(corr)
+
+    def __call__(self, coords):
+        r = self.radius
+        coords = coords.permute(0, 2, 3, 1)
+        batch, h1, w1, _ = coords.shape
+
+        out_pyramid = []
+        for i in range(self.num_levels):
+            corr = self.corr_pyramid[i]
+            dx = torch.linspace(-r, r, 2*r+1)
+            dy = torch.linspace(-r, r, 2*r+1)
+            delta = torch.stack(torch.meshgrid(dy, dx), axis=-1).to(coords.device)
+
+            centroid_lvl = coords.reshape(batch*h1*w1, 1, 1, 2) / 2**i
+            delta_lvl = delta.view(1, 2*r+1, 2*r+1, 2)
+            coords_lvl = centroid_lvl + delta_lvl
+
+            corr = bilinear_sampler(corr, coords_lvl)
+            corr = corr.view(batch, h1, w1, -1)
+            out_pyramid.append(corr)
+
+        out = torch.cat(out_pyramid, dim=-1)
+        return out.permute(0, 3, 1, 2)
+
+    @staticmethod
+    def corr(fmap1, fmap2):
+        batch, dim, ht, wd = fmap1.shape
+        fmap1 = fmap1.view(batch, dim, ht*wd)
+        fmap2 = fmap2.view(batch, dim, ht*wd)
+        
+        corr = torch.matmul(fmap1.transpose(1,2), fmap2)
+        corr = corr.view(batch, ht, wd, 1, ht, wd)
+        return corr / torch.sqrt(torch.tensor(dim).float())

core/modules/extractor.py

@@ -0,0 +1,269 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+
+class ResidualBlock(nn.Module):
+    def __init__(self, in_planes, planes, norm_fn='group', stride=1):
+        super(ResidualBlock, self).__init__()
+  
+        self.conv1 = nn.Conv2d(in_planes, planes, kernel_size=3, padding=1, stride=stride)
+        self.conv2 = nn.Conv2d(planes, planes, kernel_size=3, padding=1)
+        self.relu = nn.ReLU(inplace=True)
+
+        num_groups = planes // 8
+
+        if norm_fn == 'group':
+            self.norm1 = nn.GroupNorm(num_groups=num_groups, num_channels=planes)
+            self.norm2 = nn.GroupNorm(num_groups=num_groups, num_channels=planes)
+            if not stride == 1:
+                self.norm3 = nn.GroupNorm(num_groups=num_groups, num_channels=planes)
+        
+        elif norm_fn == 'batch':
+            self.norm1 = nn.BatchNorm2d(planes)
+            self.norm2 = nn.BatchNorm2d(planes)
+            if not stride == 1:
+                self.norm3 = nn.BatchNorm2d(planes)
+        
+        elif norm_fn == 'instance':
+            self.norm1 = nn.InstanceNorm2d(planes)
+            self.norm2 = nn.InstanceNorm2d(planes)
+            if not stride == 1:
+                self.norm3 = nn.InstanceNorm2d(planes)
+
+        elif norm_fn == 'none':
+            self.norm1 = nn.Sequential()
+            self.norm2 = nn.Sequential()
+            if not stride == 1:
+                self.norm3 = nn.Sequential()
+
+        if stride == 1:
+            self.downsample = None
+        
+        else:    
+            self.downsample = nn.Sequential(
+                nn.Conv2d(in_planes, planes, kernel_size=1, stride=stride), self.norm3)
+
+
+    def forward(self, x):
+        y = x
+        y = self.relu(self.norm1(self.conv1(y)))
+        y = self.relu(self.norm2(self.conv2(y)))
+
+        if self.downsample is not None:
+            x = self.downsample(x)
+
+        return self.relu(x+y)
+
+
+
+class BottleneckBlock(nn.Module):
+    def __init__(self, in_planes, planes, norm_fn='group', stride=1):
+        super(BottleneckBlock, self).__init__()
+  
+        self.conv1 = nn.Conv2d(in_planes, planes//4, kernel_size=1, padding=0)
+        self.conv2 = nn.Conv2d(planes//4, planes//4, kernel_size=3, padding=1, stride=stride)
+        self.conv3 = nn.Conv2d(planes//4, planes, kernel_size=1, padding=0)
+        self.relu = nn.ReLU(inplace=True)
+
+        num_groups = planes // 8
+
+        if norm_fn == 'group':
+            self.norm1 = nn.GroupNorm(num_groups=num_groups, num_channels=planes//4)
+            self.norm2 = nn.GroupNorm(num_groups=num_groups, num_channels=planes//4)
+            self.norm3 = nn.GroupNorm(num_groups=num_groups, num_channels=planes)
+            if not stride == 1:
+                self.norm4 = nn.GroupNorm(num_groups=num_groups, num_channels=planes)
+        
+        elif norm_fn == 'batch':
+            self.norm1 = nn.BatchNorm2d(planes//4)
+            self.norm2 = nn.BatchNorm2d(planes//4)
+            self.norm3 = nn.BatchNorm2d(planes)
+            if not stride == 1:
+                self.norm4 = nn.BatchNorm2d(planes)
+        
+        elif norm_fn == 'instance':
+            self.norm1 = nn.InstanceNorm2d(planes//4)
+            self.norm2 = nn.InstanceNorm2d(planes//4)
+            self.norm3 = nn.InstanceNorm2d(planes)
+            if not stride == 1:
+                self.norm4 = nn.InstanceNorm2d(planes)
+
+        elif norm_fn == 'none':
+            self.norm1 = nn.Sequential()
+            self.norm2 = nn.Sequential()
+            self.norm3 = nn.Sequential()
+            if not stride == 1:
+                self.norm4 = nn.Sequential()
+
+        if stride == 1:
+            self.downsample = None
+        
+        else:    
+            self.downsample = nn.Sequential(
+                nn.Conv2d(in_planes, planes, kernel_size=1, stride=stride), self.norm4)
+
+
+    def forward(self, x):
+        y = x
+        y = self.relu(self.norm1(self.conv1(y)))
+        y = self.relu(self.norm2(self.conv2(y)))
+        y = self.relu(self.norm3(self.conv3(y)))
+
+        if self.downsample is not None:
+            x = self.downsample(x)
+
+        return self.relu(x+y)
+
+class BasicEncoder(nn.Module):
+    def __init__(self, output_dim=128, norm_fn='batch', dropout=0.0):
+        super(BasicEncoder, self).__init__()
+        self.norm_fn = norm_fn
+
+        if self.norm_fn == 'group':
+            self.norm1 = nn.GroupNorm(num_groups=8, num_channels=64)
+            
+        elif self.norm_fn == 'batch':
+            self.norm1 = nn.BatchNorm2d(64)
+
+        elif self.norm_fn == 'instance':
+            self.norm1 = nn.InstanceNorm2d(64)
+
+        elif self.norm_fn == 'none':
+            self.norm1 = nn.Sequential()
+
+        self.conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3)
+        self.relu1 = nn.ReLU(inplace=True)
+
+        self.in_planes = 64
+        self.layer1 = self._make_layer(64,  stride=1)
+        self.layer2 = self._make_layer(96, stride=2)
+        self.layer3 = self._make_layer(128, stride=2)
+
+        # output convolution
+        self.conv2 = nn.Conv2d(128, output_dim, kernel_size=1)
+
+        if dropout > 0:
+            self.dropout = nn.Dropout2d(p=dropout)
+        else:
+            self.dropout = None
+
+        for m in self.modules():
+            if isinstance(m, nn.Conv2d):
+                nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu')
+            elif isinstance(m, (nn.BatchNorm2d, nn.InstanceNorm2d, nn.GroupNorm)):
+                if m.weight is not None:
+                    nn.init.constant_(m.weight, 1)
+                if m.bias is not None:
+                    nn.init.constant_(m.bias, 0)
+
+    def _make_layer(self, dim, stride=1):
+        layer1 = ResidualBlock(self.in_planes, dim, self.norm_fn, stride=stride)
+        layer2 = ResidualBlock(dim, dim, self.norm_fn, stride=1)
+        layers = (layer1, layer2)
+        
+        self.in_planes = dim
+        return nn.Sequential(*layers)
+
+
+    def forward(self, x):
+
+        # if input is list, combine batch dimension
+        is_list = isinstance(x, tuple) or isinstance(x, list)
+        if is_list:
+            batch_dim = x[0].shape[0]
+            x = torch.cat(x, dim=0)
+
+        x = self.conv1(x)
+        x = self.norm1(x)
+        x = self.relu1(x)
+
+        x = self.layer1(x)
+        x = self.layer2(x)
+        x = self.layer3(x)
+
+        x = self.conv2(x)
+
+        if self.dropout is not None:
+            x = self.dropout(x)
+
+        if is_list:
+            x = torch.split(x, [batch_dim, batch_dim], dim=0)
+
+        return x
+
+
+class SmallEncoder(nn.Module):
+    def __init__(self, output_dim=128, norm_fn='batch', dropout=0.0):
+        super(SmallEncoder, self).__init__()
+        self.norm_fn = norm_fn
+
+        if self.norm_fn == 'group':
+            self.norm1 = nn.GroupNorm(num_groups=8, num_channels=32)
+            
+        elif self.norm_fn == 'batch':
+            self.norm1 = nn.BatchNorm2d(32)
+
+        elif self.norm_fn == 'instance':
+            self.norm1 = nn.InstanceNorm2d(32)
+
+        elif self.norm_fn == 'none':
+            self.norm1 = nn.Sequential()
+
+        self.conv1 = nn.Conv2d(3, 32, kernel_size=7, stride=2, padding=3)
+        self.relu1 = nn.ReLU(inplace=True)
+
+        self.in_planes = 32
+        self.layer1 = self._make_layer(32,  stride=1)
+        self.layer2 = self._make_layer(64, stride=2)
+        self.layer3 = self._make_layer(96, stride=2)
+
+        if dropout > 0:
+            self.dropout = nn.Dropout2d(p=dropout)
+        else:
+            self.dropout = None
+        
+        self.conv2 = nn.Conv2d(96, output_dim, kernel_size=1)
+
+        for m in self.modules():
+            if isinstance(m, nn.Conv2d):
+                nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu')
+            elif isinstance(m, (nn.BatchNorm2d, nn.InstanceNorm2d, nn.GroupNorm)):
+                if m.weight is not None:
+                    nn.init.constant_(m.weight, 1)
+                if m.bias is not None:
+                    nn.init.constant_(m.bias, 0)
+
+    def _make_layer(self, dim, stride=1):
+        layer1 = BottleneckBlock(self.in_planes, dim, self.norm_fn, stride=stride)
+        layer2 = BottleneckBlock(dim, dim, self.norm_fn, stride=1)
+        layers = (layer1, layer2)
+    
+        self.in_planes = dim
+        return nn.Sequential(*layers)
+
+
+    def forward(self, x):
+
+        # if input is list, combine batch dimension
+        is_list = isinstance(x, tuple) or isinstance(x, list)
+        if is_list:
+            batch_dim = x[0].shape[0]
+            x = torch.cat(x, dim=0)
+
+        x = self.conv1(x)
+        x = self.norm1(x)
+        x = self.relu1(x)
+
+        x = self.layer1(x)
+        x = self.layer2(x)
+        x = self.layer3(x)
+        x = self.conv2(x)
+
+        # if self.dropout is not None:
+        #     x = self.dropout(x)
+
+        if is_list:
+            x = torch.split(x, [batch_dim, batch_dim], dim=0)
+
+        return x

core/modules/update.py

@@ -0,0 +1,169 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+# VariationalHidDropout from https://github.com/locuslab/trellisnet/tree/master/TrellisNet
+class VariationalHidDropout(nn.Module):
+    def __init__(self, dropout=0.0):
+        """
+        Hidden-to-hidden (VD-based) dropout that applies the same mask at every time step and every layer of TrellisNet
+        :param dropout: The dropout rate (0 means no dropout is applied)
+        """
+        super(VariationalHidDropout, self).__init__()
+        self.dropout = dropout
+        self.mask = None
+
+    def reset_mask(self, x):
+        dropout = self.dropout
+
+        # Dimension (N, C, L)
+        n, c, h, w = x.shape
+        m = x.data.new(n, c, 1, 1).bernoulli_(1 - dropout)
+        with torch.no_grad():
+            mask = m / (1 - dropout)
+            self.mask = mask
+        return mask
+
+    def forward(self, x):
+        if not self.training or self.dropout == 0:
+            return x
+        assert self.mask is not None, "You need to reset mask before using VariationalHidDropout"
+        return self.mask * x
+
+
+class FlowHead(nn.Module):
+    def __init__(self, input_dim=128, hidden_dim=256):
+        super(FlowHead, self).__init__()
+        self.conv1 = nn.Conv2d(input_dim, hidden_dim, 3, padding=1)
+        self.conv2 = nn.Conv2d(hidden_dim, 2, 3, padding=1)
+        self.relu = nn.ReLU(inplace=True)
+
+    def forward(self, x):
+        return self.conv2(self.relu(self.conv1(x)))
+
+
+class ConvGRU(nn.Module):
+    def __init__(self, hidden_dim=128, input_dim=192+128):
+        super(ConvGRU, self).__init__()
+        self.convz = nn.Conv2d(hidden_dim+input_dim, hidden_dim, 3, padding=1)
+        self.convr = nn.Conv2d(hidden_dim+input_dim, hidden_dim, 3, padding=1)
+        self.convq = nn.Conv2d(hidden_dim+input_dim, hidden_dim, 3, padding=1)
+
+    def forward(self, h, x):
+        hx = torch.cat([h, x], dim=1)
+
+        z = torch.sigmoid(self.convz(hx))
+        r = torch.sigmoid(self.convr(hx))
+        q = torch.tanh(self.convq(torch.cat([r*h, x], dim=1)))
+
+        h = (1-z) * h + z * q
+        return h
+
+
+class SepConvGRU(nn.Module):
+    def __init__(self, hidden_dim=128, input_dim=192+128):
+        super(SepConvGRU, self).__init__()
+        self.convz1 = nn.Conv2d(hidden_dim+input_dim, hidden_dim, (1,5), padding=(0,2))
+        self.convr1 = nn.Conv2d(hidden_dim+input_dim, hidden_dim, (1,5), padding=(0,2))
+        self.convq1 = nn.Conv2d(hidden_dim+input_dim, hidden_dim, (1,5), padding=(0,2))
+
+        self.convz2 = nn.Conv2d(hidden_dim+input_dim, hidden_dim, (5,1), padding=(2,0))
+        self.convr2 = nn.Conv2d(hidden_dim+input_dim, hidden_dim, (5,1), padding=(2,0))
+        self.convq2 = nn.Conv2d(hidden_dim+input_dim, hidden_dim, (5,1), padding=(2,0))
+
+
+    def forward(self, h, x):
+        # horizontal
+        hx = torch.cat([h, x], dim=1)
+        z = torch.sigmoid(self.convz1(hx))
+        r = torch.sigmoid(self.convr1(hx))
+        q = torch.tanh(self.convq1(torch.cat([r*h, x], dim=1)))        
+        h = (1-z) * h + z * q
+
+        # vertical
+        hx = torch.cat([h, x], dim=1)
+        z = torch.sigmoid(self.convz2(hx))
+        r = torch.sigmoid(self.convr2(hx))
+        q = torch.tanh(self.convq2(torch.cat([r*h, x], dim=1)))       
+        h = (1-z) * h + z * q
+
+        return h
+
+class SmallMotionEncoder(nn.Module):
+    def __init__(self, args):
+        super(SmallMotionEncoder, self).__init__()
+        cor_planes = args.corr_levels * (2*args.corr_radius + 1)**2
+        self.convc1 = nn.Conv2d(cor_planes, 96, 1, padding=0)
+        self.convf1 = nn.Conv2d(2, 64, 7, padding=3)
+        self.convf2 = nn.Conv2d(64, 32, 3, padding=1)
+        self.conv = nn.Conv2d(128, 80, 3, padding=1)
+
+    def forward(self, flow, corr):
+        cor = F.relu(self.convc1(corr))
+        flo = F.relu(self.convf1(flow))
+        flo = F.relu(self.convf2(flo))
+        cor_flo = torch.cat([cor, flo], dim=1)
+        out = F.relu(self.conv(cor_flo))
+        return torch.cat([out, flow], dim=1)
+
+class BasicMotionEncoder(nn.Module):
+    def __init__(self, args):
+        super(BasicMotionEncoder, self).__init__()
+        cor_planes = args.corr_levels * (2*args.corr_radius + 1)**2
+        self.convc1 = nn.Conv2d(cor_planes, 256, 1, padding=0)
+        self.convc2 = nn.Conv2d(256, 192, 3, padding=1)
+        self.convf1 = nn.Conv2d(2, 128, 7, padding=3)
+        self.convf2 = nn.Conv2d(128, 64, 3, padding=1)
+        self.conv = nn.Conv2d(64+192, 128-2, 3, padding=1)
+
+    def forward(self, flow, corr):
+        cor = F.relu(self.convc1(corr))
+        cor = F.relu(self.convc2(cor))
+        flo = F.relu(self.convf1(flow))
+        flo = F.relu(self.convf2(flo))
+
+        cor_flo = torch.cat([cor, flo], dim=1)
+        out = F.relu(self.conv(cor_flo))
+        return torch.cat([out, flow], dim=1)
+
+class SmallUpdateBlock(nn.Module):
+    def __init__(self, args, hidden_dim=96):
+        super(SmallUpdateBlock, self).__init__()
+        self.encoder = SmallMotionEncoder(args)
+        self.gru = ConvGRU(hidden_dim=hidden_dim, input_dim=82+64)
+        self.flow_head = FlowHead(hidden_dim, hidden_dim=128)
+
+    def forward(self, net, inp, corr, flow):
+        motion_features = self.encoder(flow, corr)
+        inp = torch.cat([inp, motion_features], dim=1)
+        net = self.gru(net, inp)
+        delta_flow = self.flow_head(net)
+
+        return net, delta_flow
+
+class BasicUpdateBlock(nn.Module):
+    def __init__(self, args, hidden_dim=128, input_dim=128):
+        super(BasicUpdateBlock, self).__init__()
+        self.encoder = BasicMotionEncoder(args)
+        self.gru = SepConvGRU(hidden_dim=hidden_dim, input_dim=128+hidden_dim)
+        self.flow_head = FlowHead(hidden_dim, hidden_dim=256)
+
+        self.drop_inp = VariationalHidDropout(dropout=args.dropout)
+        self.drop_net = VariationalHidDropout(dropout=args.dropout)
+
+    def reset_mask(self, net, inp):
+        self.drop_inp.reset_mask(inp)
+        self.drop_net.reset_mask(net)
+
+    def forward(self, net, inp, corr, flow):
+        motion_features = self.encoder(flow, corr)
+        inp = torch.cat([inp, motion_features], dim=1)
+
+        if self.training:
+            net = self.drop_net(net)
+            inp = self.drop_inp(inp)
+
+        net = self.gru(net, inp)
+        delta_flow = self.flow_head(net)
+
+        return net, delta_flow

core/raft.py

@@ -0,0 +1,99 @@
+import numpy as np
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+from modules.update import BasicUpdateBlock, SmallUpdateBlock
+from modules.extractor import BasicEncoder, SmallEncoder
+from modules.corr import CorrBlock
+from utils.utils import bilinear_sampler, coords_grid, upflow8
+
+
+class RAFT(nn.Module):
+    def __init__(self, args):
+        super(RAFT, self).__init__()
+        self.args = args
+
+        if args.small:
+            self.hidden_dim = hdim = 96
+            self.context_dim = cdim = 64
+            args.corr_levels = 4
+            args.corr_radius = 3
+        
+        else:
+            self.hidden_dim = hdim = 128
+            self.context_dim = cdim = 128
+            args.corr_levels = 4
+            args.corr_radius = 4
+
+        if 'dropout' not in args._get_kwargs():
+            args.dropout = 0
+
+        # feature network, context network, and update block
+        if args.small:
+            self.fnet = SmallEncoder(output_dim=128, norm_fn='instance', dropout=args.dropout)        
+            self.cnet = SmallEncoder(output_dim=hdim+cdim, norm_fn='none', dropout=args.dropout)
+            self.update_block = SmallUpdateBlock(self.args, hidden_dim=hdim)
+
+        else:
+            self.fnet = BasicEncoder(output_dim=256, norm_fn='instance', dropout=args.dropout)        
+            self.cnet = BasicEncoder(output_dim=hdim+cdim, norm_fn='batch', dropout=args.dropout)
+            self.update_block = BasicUpdateBlock(self.args, hidden_dim=hdim)
+
+    def freeze_bn(self):
+        for m in self.modules():
+            if isinstance(m, nn.BatchNorm2d):
+                m.eval()
+
+    def initialize_flow(self, img):
+        """ Flow is represented as difference between two coordinate grids flow = coords1 - coords0"""
+        N, C, H, W = img.shape
+        coords0 = coords_grid(N, H//8, W//8).to(img.device)
+        coords1 = coords_grid(N, H//8, W//8).to(img.device)
+
+        # optical flow computed as difference: flow = coords1 - coords0
+        return coords0, coords1
+
+    def forward(self, image1, image2, iters=12, flow_init=None, upsample=True):
+        """ Estimate optical flow between pair of frames """
+
+        image1 = 2 * (image1 / 255.0) - 1.0
+        image2 = 2 * (image2 / 255.0) - 1.0
+
+        hdim = self.hidden_dim
+        cdim = self.context_dim
+
+        # run the feature network
+        fmap1, fmap2 = self.fnet([image1, image2])
+        corr_fn = CorrBlock(fmap1, fmap2, radius=self.args.corr_radius)
+
+        # run the context network
+        cnet = self.cnet(image1)
+        net, inp = torch.split(cnet, [hdim, cdim], dim=1)
+        net, inp = torch.tanh(net), torch.relu(inp)
+
+        # if dropout is being used reset mask
+        self.update_block.reset_mask(net, inp)
+        coords0, coords1 = self.initialize_flow(image1)
+
+        flow_predictions = []
+        for itr in range(iters):
+            coords1 = coords1.detach()
+            corr = corr_fn(coords1) # index correlation volume
+
+            flow = coords1 - coords0
+            net, delta_flow = self.update_block(net, inp, corr, flow)
+
+            # F(t+1) = F(t) + \Delta(t)
+            coords1 = coords1 + delta_flow
+            
+            if upsample:
+                flow_up = upflow8(coords1 - coords0)
+                flow_predictions.append(flow_up)
+            
+            else:
+                flow_predictions.append(coords1 - coords0)
+
+        return flow_predictions
+
+

core/utils/augmentor.py

@@ -0,0 +1,233 @@
+import numpy as np
+import random
+import math
+import cv2
+from PIL import Image
+
+import torch
+import torchvision
+import torch.nn.functional as F
+
+class FlowAugmentor:
+    def __init__(self, crop_size, min_scale=-0.2, max_scale=0.5):
+        self.crop_size = crop_size
+        self.augcolor = torchvision.transforms.ColorJitter(
+            brightness=0.4, 
+            contrast=0.4, 
+            saturation=0.4, 
+            hue=0.5/3.14)
+
+        self.asymmetric_color_aug_prob = 0.2
+        self.spatial_aug_prob = 0.8
+        self.eraser_aug_prob = 0.5
+
+        self.min_scale = min_scale
+        self.max_scale = max_scale
+        self.max_stretch = 0.2
+        self.stretch_prob = 0.8
+        self.margin = 20
+
+    def color_transform(self, img1, img2):
+
+        if np.random.rand() < self.asymmetric_color_aug_prob:
+            img1 = np.array(self.augcolor(Image.fromarray(img1)), dtype=np.uint8)
+            img2 = np.array(self.augcolor(Image.fromarray(img2)), dtype=np.uint8)
+
+        else:
+            image_stack = np.concatenate([img1, img2], axis=0)
+            image_stack = np.array(self.augcolor(Image.fromarray(image_stack)), dtype=np.uint8)
+            img1, img2 = np.split(image_stack, 2, axis=0)
+
+        return img1, img2
+
+    def eraser_transform(self, img1, img2, bounds=[50, 100]):
+        ht, wd = img1.shape[:2]
+        if np.random.rand() < self.eraser_aug_prob:
+            mean_color = np.mean(img2.reshape(-1, 3), axis=0)
+            for _ in range(np.random.randint(1, 3)):
+                x0 = np.random.randint(0, wd)
+                y0 = np.random.randint(0, ht)
+                dx = np.random.randint(bounds[0], bounds[1])
+                dy = np.random.randint(bounds[0], bounds[1])
+                img2[y0:y0+dy, x0:x0+dx, :] = mean_color
+
+        return img1, img2
+
+    def spatial_transform(self, img1, img2, flow):
+        # randomly sample scale
+
+        ht, wd = img1.shape[:2]
+        min_scale = np.maximum(
+            (self.crop_size[0] + 1) / float(ht), 
+            (self.crop_size[1] + 1) / float(wd))
+
+        max_scale = self.max_scale
+        min_scale = max(min_scale, self.min_scale)
+        
+        scale = 2 ** np.random.uniform(self.min_scale, self.max_scale)
+        scale_x = scale
+        scale_y = scale
+        if np.random.rand() < self.stretch_prob:
+            scale_x *= 2 ** np.random.uniform(-self.max_stretch, self.max_stretch)
+            scale_y *= 2 ** np.random.uniform(-self.max_stretch, self.max_stretch)
+            
+        scale_x = np.clip(scale_x, min_scale, None)
+        scale_y = np.clip(scale_y, min_scale, None)
+
+        if np.random.rand() < self.spatial_aug_prob:
+            # rescale the images
+            img1 = cv2.resize(img1, None, fx=scale_x, fy=scale_y, interpolation=cv2.INTER_LINEAR)
+            img2 = cv2.resize(img2, None, fx=scale_x, fy=scale_y, interpolation=cv2.INTER_LINEAR)
+            flow = cv2.resize(flow, None, fx=scale_x, fy=scale_y, interpolation=cv2.INTER_LINEAR)
+            flow = flow * [scale_x, scale_y]
+
+        if np.random.rand() < 0.5: # h-flip
+            img1 = img1[:, ::-1]
+            img2 = img2[:, ::-1]
+            flow = flow[:, ::-1] * [-1.0, 1.0]
+
+        if np.random.rand() < 0.1: # v-flip
+            img1 = img1[::-1, :]
+            img2 = img2[::-1, :]
+            flow = flow[::-1, :] * [1.0, -1.0]
+
+        y0 = np.random.randint(-self.margin, img1.shape[0] - self.crop_size[0] + self.margin)
+        x0 = np.random.randint(-self.margin, img1.shape[1] - self.crop_size[1] + self.margin)
+        
+        y0 = np.clip(y0, 0, img1.shape[0] - self.crop_size[0])
+        x0 = np.clip(x0, 0, img1.shape[1] - self.crop_size[1])
+
+        img1 = img1[y0:y0+self.crop_size[0], x0:x0+self.crop_size[1]]
+        img2 = img2[y0:y0+self.crop_size[0], x0:x0+self.crop_size[1]]
+        flow = flow[y0:y0+self.crop_size[0], x0:x0+self.crop_size[1]]
+
+        return img1, img2, flow
+
+    def __call__(self, img1, img2, flow):
+        img1, img2 = self.color_transform(img1, img2)
+        img1, img2 = self.eraser_transform(img1, img2)
+        img1, img2, flow = self.spatial_transform(img1, img2, flow)
+
+        img1 = np.ascontiguousarray(img1)
+        img2 = np.ascontiguousarray(img2)
+        flow = np.ascontiguousarray(flow)
+
+        return img1, img2, flow
+
+
+class FlowAugmentorKITTI:
+    def __init__(self, crop_size, min_scale=-0.2, max_scale=0.5):
+        self.crop_size = crop_size
+        self.augcolor = torchvision.transforms.ColorJitter(
+            brightness=0.3, contrast=0.3, saturation=0.3, hue=0.3/3.14)
+
+        self.max_scale = max_scale
+        self.min_scale = min_scale
+
+        self.spatial_aug_prob = 0.8
+        self.eraser_aug_prob = 0.5
+        
+    def color_transform(self, img1, img2):
+        image_stack = np.concatenate([img1, img2], axis=0)
+        image_stack = np.array(self.augcolor(Image.fromarray(image_stack)), dtype=np.uint8)
+        img1, img2 = np.split(image_stack, 2, axis=0)
+        return img1, img2
+
+    def eraser_transform(self, img1, img2):
+        ht, wd = img1.shape[:2]
+        if np.random.rand() < self.eraser_aug_prob:
+            mean_color = np.mean(img2.reshape(-1, 3), axis=0)
+            for _ in range(np.random.randint(1, 3)):
+                x0 = np.random.randint(0, wd)
+                y0 = np.random.randint(0, ht)
+                dx = np.random.randint(50, 100)
+                dy = np.random.randint(50, 100)
+                img2[y0:y0+dy, x0:x0+dx, :] = mean_color
+
+        return img1, img2
+
+    def resize_sparse_flow_map(self, flow, valid, fx=1.0, fy=1.0):
+        ht, wd = flow.shape[:2]
+        coords = np.meshgrid(np.arange(wd), np.arange(ht))
+        coords = np.stack(coords, axis=-1)
+
+        coords = coords.reshape(-1, 2).astype(np.float32)
+        flow = flow.reshape(-1, 2).astype(np.float32)
+        valid = valid.reshape(-1).astype(np.float32)
+
+        coords0 = coords[valid>=1]
+        flow0 = flow[valid>=1]
+
+        ht1 = int(round(ht * fy))
+        wd1 = int(round(wd * fx))
+
+        coords1 = coords0 * [fx, fy]
+        flow1 = flow0 * [fx, fy]
+
+        xx = np.round(coords1[:,0]).astype(np.int32)
+        yy = np.round(coords1[:,1]).astype(np.int32)
+
+        v = (xx > 0) & (xx < wd1) & (yy > 0) & (yy < ht1)
+        xx = xx[v]
+        yy = yy[v]
+        flow1 = flow1[v]
+
+        flow_img = np.zeros([ht1, wd1, 2], dtype=np.float32)
+        valid_img = np.zeros([ht1, wd1], dtype=np.int32)
+
+        flow_img[yy, xx] = flow1
+        valid_img[yy, xx] = 1
+
+        return flow_img, valid_img
+
+    def spatial_transform(self, img1, img2, flow, valid):
+        # randomly sample scale
+
+        ht, wd = img1.shape[:2]
+        min_scale = np.maximum(
+            (self.crop_size[0] + 1) / float(ht), 
+            (self.crop_size[1] + 1) / float(wd))
+
+        scale = 2 ** np.random.uniform(self.min_scale, self.max_scale)
+        scale_x = np.clip(scale, min_scale, None)
+        scale_y = np.clip(scale, min_scale, None)
+
+        if np.random.rand() < self.spatial_aug_prob:
+            # rescale the images
+            img1 = cv2.resize(img1, None, fx=scale_x, fy=scale_y, interpolation=cv2.INTER_LINEAR)
+            img2 = cv2.resize(img2, None, fx=scale_x, fy=scale_y, interpolation=cv2.INTER_LINEAR)
+            flow, valid = self.resize_sparse_flow_map(flow, valid, fx=scale_x, fy=scale_y)
+
+        if np.random.rand() < 0.5: # h-flip
+            img1 = img1[:, ::-1]
+            img2 = img2[:, ::-1]
+            flow = flow[:, ::-1] * [-1.0, 1.0]
+            valid = valid[:, ::-1]
+
+        margin_y = 20
+        margin_x = 50
+
+        y0 = np.random.randint(0, img1.shape[0] - self.crop_size[0] + margin_y)
+        x0 = np.random.randint(-margin_x, img1.shape[1] - self.crop_size[1] + margin_x)
+
+        y0 = np.clip(y0, 0, img1.shape[0] - self.crop_size[0])
+        x0 = np.clip(x0, 0, img1.shape[1] - self.crop_size[1])
+
+        img1 = img1[y0:y0+self.crop_size[0], x0:x0+self.crop_size[1]]
+        img2 = img2[y0:y0+self.crop_size[0], x0:x0+self.crop_size[1]]
+        flow = flow[y0:y0+self.crop_size[0], x0:x0+self.crop_size[1]]
+        valid = valid[y0:y0+self.crop_size[0], x0:x0+self.crop_size[1]]
+        return img1, img2, flow, valid
+
+
+    def __call__(self, img1, img2, flow, valid):
+        img1, img2 = self.color_transform(img1, img2)
+        img1, img2 = self.eraser_transform(img1, img2)
+        img1, img2, flow, valid = self.spatial_transform(img1, img2, flow, valid)
+
+        img1 = np.ascontiguousarray(img1)
+        img2 = np.ascontiguousarray(img2)
+        flow = np.ascontiguousarray(flow)
+        valid = np.ascontiguousarray(valid)
+
+        return img1, img2, flow, valid

core/utils/flow_viz.py

@@ -0,0 +1,275 @@
+# MIT License
+#
+# Copyright (c) 2018 Tom Runia
+#
+# Permission is hereby granted, free of charge, to any person obtaining a copy
+# of this software and associated documentation files (the "Software"), to deal
+# in the Software without restriction, including without limitation the rights
+# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+# copies of the Software, and to permit persons to whom the Software is
+# furnished to do so, subject to conditions.
+#
+# Author: Tom Runia
+# Date Created: 2018-08-03
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+import numpy as np
+
+
+def make_colorwheel():
+    '''
+    Generates a color wheel for optical flow visualization as presented in:
+        Baker et al. "A Database and Evaluation Methodology for Optical Flow" (ICCV, 2007)
+        URL: http://vision.middlebury.edu/flow/flowEval-iccv07.pdf
+    According to the C++ source code of Daniel Scharstein
+    According to the Matlab source code of Deqing Sun
+    '''
+
+    RY = 15
+    YG = 6
+    GC = 4
+    CB = 11
+    BM = 13
+    MR = 6
+
+    ncols = RY + YG + GC + CB + BM + MR
+    colorwheel = np.zeros((ncols, 3))
+    col = 0
+
+    # RY
+    colorwheel[0:RY, 0] = 255
+    colorwheel[0:RY, 1] = np.floor(255*np.arange(0,RY)/RY)
+    col = col+RY
+    # YG
+    colorwheel[col:col+YG, 0] = 255 - np.floor(255*np.arange(0,YG)/YG)
+    colorwheel[col:col+YG, 1] = 255
+    col = col+YG
+    # GC
+    colorwheel[col:col+GC, 1] = 255
+    colorwheel[col:col+GC, 2] = np.floor(255*np.arange(0,GC)/GC)
+    col = col+GC
+    # CB
+    colorwheel[col:col+CB, 1] = 255 - np.floor(255*np.arange(CB)/CB)
+    colorwheel[col:col+CB, 2] = 255
+    col = col+CB
+    # BM
+    colorwheel[col:col+BM, 2] = 255
+    colorwheel[col:col+BM, 0] = np.floor(255*np.arange(0,BM)/BM)
+    col = col+BM
+    # MR
+    colorwheel[col:col+MR, 2] = 255 - np.floor(255*np.arange(MR)/MR)
+    colorwheel[col:col+MR, 0] = 255
+    return colorwheel
+
+
+def flow_compute_color(u, v, convert_to_bgr=False):
+    '''
+    Applies the flow color wheel to (possibly clipped) flow components u and v.
+    According to the C++ source code of Daniel Scharstein
+    According to the Matlab source code of Deqing Sun
+    :param u: np.ndarray, input horizontal flow
+    :param v: np.ndarray, input vertical flow
+    :param convert_to_bgr: bool, whether to change ordering and output BGR instead of RGB
+    :return:
+    '''
+
+    flow_image = np.zeros((u.shape[0], u.shape[1], 3), np.uint8)
+
+    colorwheel = make_colorwheel()  # shape [55x3]
+    ncols = colorwheel.shape[0]
+
+    rad = np.sqrt(np.square(u) + np.square(v))
+    a = np.arctan2(-v, -u)/np.pi
+
+    fk = (a+1) / 2*(ncols-1) + 1
+    k0 = np.floor(fk).astype(np.int32)
+    k1 = k0 + 1
+    k1[k1 == ncols] = 1
+    f = fk - k0
+
+    for i in range(colorwheel.shape[1]):
+
+        tmp = colorwheel[:,i]
+        col0 = tmp[k0] / 255.0
+        col1 = tmp[k1] / 255.0
+        col = (1-f)*col0 + f*col1
+
+        idx = (rad <= 1)
+        col[idx]  = 1 - rad[idx] * (1-col[idx])
+        col[~idx] = col[~idx] * 0.75   # out of range?
+
+        # Note the 2-i => BGR instead of RGB
+        ch_idx = 2-i if convert_to_bgr else i
+        flow_image[:,:,ch_idx] = np.floor(255 * col)
+
+    return flow_image
+
+
+def flow_to_color(flow_uv, clip_flow=None, convert_to_bgr=False):
+    '''
+    Expects a two dimensional flow image of shape [H,W,2]
+    According to the C++ source code of Daniel Scharstein
+    According to the Matlab source code of Deqing Sun
+    :param flow_uv: np.ndarray of shape [H,W,2]
+    :param clip_flow: float, maximum clipping value for flow
+    :return:
+    '''
+
+    assert flow_uv.ndim == 3, 'input flow must have three dimensions'
+    assert flow_uv.shape[2] == 2, 'input flow must have shape [H,W,2]'
+
+    if clip_flow is not None:
+        flow_uv = np.clip(flow_uv, 0, clip_flow)
+
+    u = flow_uv[:,:,0]
+    v = flow_uv[:,:,1]
+
+    rad = np.sqrt(np.square(u) + np.square(v))
+    rad_max = np.max(rad)
+
+    epsilon = 1e-5
+    u = u / (rad_max + epsilon)
+    v = v / (rad_max + epsilon)
+
+    return flow_compute_color(u, v, convert_to_bgr)
+
+
+
+UNKNOWN_FLOW_THRESH = 1e7
+SMALLFLOW = 0.0
+LARGEFLOW = 1e8
+
+def make_color_wheel():
+    """
+    Generate color wheel according Middlebury color code
+    :return: Color wheel
+    """
+    RY = 15
+    YG = 6
+    GC = 4
+    CB = 11
+    BM = 13
+    MR = 6
+
+    ncols = RY + YG + GC + CB + BM + MR
+
+    colorwheel = np.zeros([ncols, 3])
+
+    col = 0
+
+    # RY
+    colorwheel[0:RY, 0] = 255
+    colorwheel[0:RY, 1] = np.transpose(np.floor(255*np.arange(0, RY) / RY))
+    col += RY
+
+    # YG
+    colorwheel[col:col+YG, 0] = 255 - np.transpose(np.floor(255*np.arange(0, YG) / YG))
+    colorwheel[col:col+YG, 1] = 255
+    col += YG
+
+    # GC
+    colorwheel[col:col+GC, 1] = 255
+    colorwheel[col:col+GC, 2] = np.transpose(np.floor(255*np.arange(0, GC) / GC))
+    col += GC
+
+    # CB
+    colorwheel[col:col+CB, 1] = 255 - np.transpose(np.floor(255*np.arange(0, CB) / CB))
+    colorwheel[col:col+CB, 2] = 255
+    col += CB
+
+    # BM
+    colorwheel[col:col+BM, 2] = 255
+    colorwheel[col:col+BM, 0] = np.transpose(np.floor(255*np.arange(0, BM) / BM))
+    col += + BM
+
+    # MR
+    colorwheel[col:col+MR, 2] = 255 - np.transpose(np.floor(255 * np.arange(0, MR) / MR))
+    colorwheel[col:col+MR, 0] = 255
+
+    return colorwheel
+
+
+
+def compute_color(u, v):
+    """
+    compute optical flow color map
+    :param u: optical flow horizontal map
+    :param v: optical flow vertical map
+    :return: optical flow in color code
+    """
+    [h, w] = u.shape
+    img = np.zeros([h, w, 3])
+    nanIdx = np.isnan(u) | np.isnan(v)
+    u[nanIdx] = 0
+    v[nanIdx] = 0
+
+    colorwheel = make_color_wheel()
+    ncols = np.size(colorwheel, 0)
+
+    rad = np.sqrt(u**2+v**2)
+
+    a = np.arctan2(-v, -u) / np.pi
+
+    fk = (a+1) / 2 * (ncols - 1) + 1
+
+    k0 = np.floor(fk).astype(int)
+
+    k1 = k0 + 1
+    k1[k1 == ncols+1] = 1
+    f = fk - k0
+
+    for i in range(0, np.size(colorwheel,1)):
+        tmp = colorwheel[:, i]
+        col0 = tmp[k0-1] / 255
+        col1 = tmp[k1-1] / 255
+        col = (1-f) * col0 + f * col1
+
+        idx = rad <= 1
+        col[idx] = 1-rad[idx]*(1-col[idx])
+        notidx = np.logical_not(idx)
+
+        col[notidx] *= 0.75
+        img[:, :, i] = np.uint8(np.floor(255 * col*(1-nanIdx)))
+
+    return img
+
+# from https://github.com/gengshan-y/VCN
+def flow_to_image(flow):
+    """
+    Convert flow into middlebury color code image
+    :param flow: optical flow map
+    :return: optical flow image in middlebury color
+    """
+    u = flow[:, :, 0]
+    v = flow[:, :, 1]
+
+    maxu = -999.
+    maxv = -999.
+    minu = 999.
+    minv = 999.
+
+    idxUnknow = (abs(u) > UNKNOWN_FLOW_THRESH) | (abs(v) > UNKNOWN_FLOW_THRESH)
+    u[idxUnknow] = 0
+    v[idxUnknow] = 0
+
+    maxu = max(maxu, np.max(u))
+    minu = min(minu, np.min(u))
+
+    maxv = max(maxv, np.max(v))
+    minv = min(minv, np.min(v))
+
+    rad = np.sqrt(u ** 2 + v ** 2)
+    maxrad = max(-1, np.max(rad))
+
+    u = u/(maxrad + np.finfo(float).eps)
+    v = v/(maxrad + np.finfo(float).eps)
+
+    img = compute_color(u, v)
+
+    idx = np.repeat(idxUnknow[:, :, np.newaxis], 3, axis=2)
+    img[idx] = 0
+
+    return np.uint8(img)

core/utils/frame_utils.py

@@ -0,0 +1,124 @@
+import numpy as np
+from PIL import Image
+from os.path import *
+import re
+import cv2
+
+TAG_CHAR = np.array([202021.25], np.float32)
+
+def readFlow(fn):
+    """ Read .flo file in Middlebury format"""
+    # Code adapted from:
+    # http://stackoverflow.com/questions/28013200/reading-middlebury-flow-files-with-python-bytes-array-numpy
+
+    # WARNING: this will work on little-endian architectures (eg Intel x86) only!
+    # print 'fn = %s'%(fn)
+    with open(fn, 'rb') as f:
+        magic = np.fromfile(f, np.float32, count=1)
+        if 202021.25 != magic:
+            print('Magic number incorrect. Invalid .flo file')
+            return None
+        else:
+            w = np.fromfile(f, np.int32, count=1)
+            h = np.fromfile(f, np.int32, count=1)
+            # print 'Reading %d x %d flo file\n' % (w, h)
+            data = np.fromfile(f, np.float32, count=2*int(w)*int(h))
+            # Reshape data into 3D array (columns, rows, bands)
+            # The reshape here is for visualization, the original code is (w,h,2)
+            return np.resize(data, (int(h), int(w), 2))
+
+def readPFM(file):
+    file = open(file, 'rb')
+
+    color = None
+    width = None
+    height = None
+    scale = None
+    endian = None
+
+    header = file.readline().rstrip()
+    if header == b'PF':
+        color = True
+    elif header == b'Pf':
+        color = False
+    else:
+        raise Exception('Not a PFM file.')
+
+    dim_match = re.match(rb'^(\d+)\s(\d+)\s$', file.readline())
+    if dim_match:
+        width, height = map(int, dim_match.groups())
+    else:
+        raise Exception('Malformed PFM header.')
+
+    scale = float(file.readline().rstrip())
+    if scale < 0: # little-endian
+        endian = '<'
+        scale = -scale
+    else:
+        endian = '>' # big-endian
+
+    data = np.fromfile(file, endian + 'f')
+    shape = (height, width, 3) if color else (height, width)
+
+    data = np.reshape(data, shape)
+    data = np.flipud(data)
+    return data
+
+def writeFlow(filename,uv,v=None):
+    """ Write optical flow to file.
+    
+    If v is None, uv is assumed to contain both u and v channels,
+    stacked in depth.
+    Original code by Deqing Sun, adapted from Daniel Scharstein.
+    """
+    nBands = 2
+
+    if v is None:
+        assert(uv.ndim == 3)
+        assert(uv.shape[2] == 2)
+        u = uv[:,:,0]
+        v = uv[:,:,1]
+    else:
+        u = uv
+
+    assert(u.shape == v.shape)
+    height,width = u.shape
+    f = open(filename,'wb')
+    # write the header
+    f.write(TAG_CHAR)
+    np.array(width).astype(np.int32).tofile(f)
+    np.array(height).astype(np.int32).tofile(f)
+    # arrange into matrix form
+    tmp = np.zeros((height, width*nBands))
+    tmp[:,np.arange(width)*2] = u
+    tmp[:,np.arange(width)*2 + 1] = v
+    tmp.astype(np.float32).tofile(f)
+    f.close()
+
+
+def readFlowKITTI(filename):
+    flow = cv2.imread(filename, cv2.IMREAD_ANYDEPTH|cv2.IMREAD_COLOR)
+    flow = flow[:,:,::-1].astype(np.float32)
+    flow, valid = flow[:, :, :2], flow[:, :, 2]
+    flow = (flow - 2**15) / 64.0
+    return flow, valid
+
+def writeFlowKITTI(filename, uv):
+    uv = 64.0 * uv + 2**15
+    valid = np.ones([uv.shape[0], uv.shape[1], 1])
+    uv = np.concatenate([uv, valid], axis=-1).astype(np.uint16)
+    cv2.imwrite(filename, uv[..., ::-1])
+    
+
+def read_gen(file_name, pil=False):
+    ext = splitext(file_name)[-1]
+    if ext == '.png' or ext == '.jpeg' or ext == '.ppm' or ext == '.jpg':
+        return Image.open(file_name)
+    elif ext == '.bin' or ext == '.raw':
+        return np.load(file_name)
+    elif ext == '.flo':
+        return readFlow(file_name).astype(np.float32)
+    elif ext == '.pfm':
+        flow = readPFM(file_name).astype(np.float32)
+        return flow[:, :, :-1]
+    return []

core/utils/utils.py

@@ -0,0 +1,62 @@
+import torch
+import torch.nn.functional as F
+import numpy as np
+from scipy import interpolate
+
+
+def bilinear_sampler(img, coords, mode='bilinear', mask=False):
+    """ Wrapper for grid_sample, uses pixel coordinates """
+    H, W = img.shape[-2:]
+    xgrid, ygrid = coords.split([1,1], dim=-1)
+    xgrid = 2*xgrid/(W-1) - 1
+    ygrid = 2*ygrid/(H-1) - 1
+
+    grid = torch.cat([xgrid, ygrid], dim=-1)
+    img = F.grid_sample(img, grid, align_corners=True)
+
+    if mask:
+        mask = (xgrid > -1) & (ygrid > -1) & (xgrid < 1) & (ygrid < 1)
+        return img, mask.float()
+
+    return img
+
+def forward_interpolate(flow):
+    flow = flow.detach().cpu().numpy()
+    dx, dy = flow[0], flow[1]
+
+    ht, wd = dx.shape
+    x0, y0 = np.meshgrid(np.arange(wd), np.arange(ht))
+
+    x1 = x0 + dx
+    y1 = y0 + dy
+    
+    x1 = x1.reshape(-1)
+    y1 = y1.reshape(-1)
+    dx = dx.reshape(-1)
+    dy = dy.reshape(-1)
+
+    valid = (x1 > 0) & (x1 < wd) & (y1 > 0) & (y1 < ht)
+    x1 = x1[valid]
+    y1 = y1[valid]
+    dx = dx[valid]
+    dy = dy[valid]
+
+    flow_x = interpolate.griddata(
+        (x1, y1), dx, (x0, y0), method='nearest')
+
+    flow_y = interpolate.griddata(
+        (x1, y1), dy, (x0, y0), method='nearest')
+
+    flow = np.stack([flow_x, flow_y], axis=0)
+    return torch.from_numpy(flow).float()
+
+
+def coords_grid(batch, ht, wd):
+    coords = torch.meshgrid(torch.arange(ht), torch.arange(wd))
+    coords = torch.stack(coords[::-1], dim=0).float()
+    return coords[None].repeat(batch, 1, 1, 1)
+
+
+def upflow8(flow, mode='bilinear'):
+    new_size = (8 * flow.shape[2], 8 * flow.shape[3])
+    return  8 * F.interpolate(flow, size=new_size, mode=mode, align_corners=True)

demo.py

@@ -0,0 +1,90 @@
+import sys
+sys.path.append('core')
+
+import argparse
+import os
+import cv2
+import numpy as np
+import torch
+import torch.nn.functional as F
+from PIL import Image
+
+import datasets
+from utils import flow_viz
+from raft import RAFT
+
+
+DEVICE = 'cuda'
+
+def pad8(img):
+    """pad image such that dimensions are divisible by 8"""
+    ht, wd = img.shape[2:]
+    pad_ht = (((ht // 8) + 1) * 8 - ht) % 8
+    pad_wd = (((wd // 8) + 1) * 8 - wd) % 8
+    pad_ht1 = [pad_ht//2, pad_ht-pad_ht//2]
+    pad_wd1 = [pad_wd//2, pad_wd-pad_wd//2]
+
+    img = F.pad(img, pad_wd1 + pad_ht1, mode='replicate')
+    return img
+
+def load_image(imfile):
+    img = np.array(Image.open(imfile)).astype(np.uint8)[..., :3]
+    img = torch.from_numpy(img).permute(2, 0, 1).float()
+    return pad8(img[None]).to(DEVICE)
+
+
+def display(image1, image2, flow):
+    image1 = image1.permute(1, 2, 0).cpu().numpy() / 255.0
+    image2 = image2.permute(1, 2, 0).cpu().numpy() / 255.0
+
+    flow = flow.permute(1, 2, 0).cpu().numpy()
+    flow_image = flow_viz.flow_to_image(flow)
+    flow_image = cv2.resize(flow_image, (image1.shape[1], image1.shape[0]))
+
+
+    cv2.imshow('image1', image1[..., ::-1])
+    cv2.imshow('image2', image2[..., ::-1])
+    cv2.imshow('flow', flow_image[..., ::-1])
+    cv2.waitKey()
+
+
+def demo(args):
+    model = RAFT(args)
+    model = torch.nn.DataParallel(model)
+    model.load_state_dict(torch.load(args.model))
+
+    model.to(DEVICE)
+    model.eval()
+
+    with torch.no_grad():
+
+        # sintel images
+        image1 = load_image('images/sintel_0.png')
+        image2 = load_image('images/sintel_1.png')
+
+        flow_predictions = model(image1, image2, iters=args.iters, upsample=False)
+        display(image1[0], image2[0], flow_predictions[-1][0])
+
+        # kitti images
+        image1 = load_image('images/kitti_0.png')
+        image2 = load_image('images/kitti_1.png')
+
+        flow_predictions = model(image1, image2, iters=16)    
+        display(image1[0], image2[0], flow_predictions[-1][0])
+
+        # davis images
+        image1 = load_image('images/davis_0.jpg')
+        image2 = load_image('images/davis_1.jpg')
+
+        flow_predictions = model(image1, image2, iters=16)    
+        display(image1[0], image2[0], flow_predictions[-1][0])
+
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--model', help="restore checkpoint")
+    parser.add_argument('--small', action='store_true', help='use small model')
+    parser.add_argument('--iters', type=int, default=12)
+
+    args = parser.parse_args()
+    demo(args)

evaluate.py

@@ -0,0 +1,100 @@
+import sys
+sys.path.append('core')
+
+from PIL import Image
+import cv2
+import argparse
+import os
+import time
+import numpy as np
+import torch
+import torch.nn.functional as F
+import matplotlib.pyplot as plt
+
+import datasets
+from utils import flow_viz
+from raft import RAFT
+
+    
+
+def validate_sintel(args, model, iters=50):
+    """ Evaluate trained model on Sintel(train) clean + final passes """
+    model.eval()
+    pad = 2
+
+    for dstype in ['clean', 'final']:
+        val_dataset = datasets.MpiSintel(args, do_augument=False, dstype=dstype)
+        
+        epe_list = []
+        for i in range(len(val_dataset)):
+            image1, image2, flow_gt, _ = val_dataset[i]
+            image1 = image1[None].cuda()
+            image2 = image2[None].cuda()
+            image1 = F.pad(image1, [0, 0, pad, pad], mode='replicate')
+            image2 = F.pad(image2, [0, 0, pad, pad], mode='replicate')
+
+            with torch.no_grad():
+                flow_predictions = model.module(image1, image2, iters=iters)
+                flow_pr = flow_predictions[-1][0,:,pad:-pad]
+
+            epe = torch.sum((flow_pr - flow_gt.cuda())**2, dim=0)
+            epe = torch.sqrt(epe).mean()
+            epe_list.append(epe.item())
+
+        print("Validation (%s) EPE: %f" % (dstype, np.mean(epe_list)))
+
+
+def validate_kitti(args, model, iters=32):
+    """ Evaluate trained model on KITTI (train) """
+
+    model.eval()
+    val_dataset = datasets.KITTI(args, do_augument=False, is_val=True, do_pad=True)
+
+    with torch.no_grad():
+        epe_list, out_list = [], []
+        for i in range(len(val_dataset)):
+            image1, image2, flow_gt, valid_gt = val_dataset[i]
+            image1 = image1[None].cuda()
+            image2 = image2[None].cuda()
+            flow_gt = flow_gt.cuda()
+            valid_gt = valid_gt.cuda()
+
+            flow_predictions = model.module(image1, image2, iters=iters)
+            flow_pr = flow_predictions[-1][0]
+
+            epe = torch.sum((flow_pr - flow_gt)**2, dim=0).sqrt()
+            mag = torch.sum(flow_gt**2, dim=0).sqrt()
+            
+            epe = epe.view(-1)
+            mag = mag.view(-1)
+            val = valid_gt.view(-1) >= 0.5
+
+            out = ((epe > 3.0) & ((epe/mag) > 0.05)).float()
+            epe_list.append(epe[val].mean().item())
+            out_list.append(out[val].cpu().numpy())
+
+    epe_list = np.array(epe_list)
+    out_list = np.concatenate(out_list)
+
+
+    print("Validation KITTI: %f, %f" % (np.mean(epe_list), 100*np.mean(out_list)))
+
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--model', help="restore checkpoint")
+    parser.add_argument('--small', action='store_true', help='use small model')
+    parser.add_argument('--sintel_iters', type=int, default=50)
+    parser.add_argument('--kitti_iters', type=int, default=32)
+
+    args = parser.parse_args()
+
+    model = RAFT(args)
+    model = torch.nn.DataParallel(model)
+    model.load_state_dict(torch.load(args.model))
+
+    model.to('cuda')
+    model.eval()
+
+    validate_sintel(args, model, args.sintel_iters)
+    validate_kitti(args, model, args.kitti_iters)

scripts/download_models.sh

@@ -0,0 +1,3 @@
+#!/bin/bash
+wget https://www.dropbox.com/s/a2acvmczgzm6f9n/models.zip
+unzip models.zip

train.py

@@ -0,0 +1,211 @@
+from __future__ import print_function, division
+import sys
+sys.path.append('core')
+
+import argparse
+import os
+import cv2
+import time
+import numpy as np
+import matplotlib.pyplot as plt
+
+import torch
+import torch.nn as nn
+import torch.optim as optim
+import torch.nn.functional as F
+
+from torch.utils.data import DataLoader
+from raft import RAFT
+from evaluate import validate_sintel, validate_kitti
+import datasets
+
+# exclude extremly large displacements
+MAX_FLOW = 1000
+SUM_FREQ = 1000
+VAL_FREQ = 5000
+
+
+def count_parameters(model):
+    return sum(p.numel() for p in model.parameters() if p.requires_grad)
+
+def sequence_loss(flow_preds, flow_gt, valid):
+    """ Loss function defined over sequence of flow predictions """
+
+    n_predictions = len(flow_preds)    
+    flow_loss = 0.0
+
+    # exlude invalid pixels and extremely large diplacements
+    valid = (valid >= 0.5) & (flow_gt.abs().sum(dim=1) < MAX_FLOW)
+
+    for i in range(n_predictions):
+        i_weight = 0.8**(n_predictions - i - 1)
+        i_loss = (flow_preds[i] - flow_gt).abs()
+        flow_loss += i_weight * (valid[:, None] * i_loss).mean()
+
+    epe = torch.sum((flow_preds[-1] - flow_gt)**2, dim=1).sqrt()
+    epe = epe.view(-1)[valid.view(-1)]
+
+    metrics = {
+        'epe': epe.mean().item(),
+        '1px': (epe < 1).float().mean().item(),
+        '3px': (epe < 3).float().mean().item(),
+        '5px': (epe < 5).float().mean().item(),
+    }
+
+    return flow_loss, metrics
+
+
+def fetch_dataloader(args):
+    """ Create the data loader for the corresponding trainign set """
+
+    if args.dataset == 'chairs':
+        train_dataset = datasets.FlyingChairs(args, image_size=args.image_size)
+    
+    elif args.dataset == 'things':
+        clean_dataset = datasets.SceneFlow(args, image_size=args.image_size, dstype='frames_cleanpass')
+        final_dataset = datasets.SceneFlow(args, image_size=args.image_size, dstype='frames_finalpass')
+        train_dataset = clean_dataset + final_dataset
+
+    elif args.dataset == 'sintel':
+        clean_dataset = datasets.MpiSintel(args, image_size=args.image_size, dstype='clean')
+        final_dataset = datasets.MpiSintel(args, image_size=args.image_size, dstype='final')
+        train_dataset = clean_dataset + final_dataset
+
+    elif args.dataset == 'kitti':
+        train_dataset = datasets.KITTI(args, image_size=args.image_size, is_val=False)
+
+    gpuargs = {'num_workers': 4, 'drop_last' : True}
+    train_loader = DataLoader(train_dataset, batch_size=args.batch_size, 
+        pin_memory=True, shuffle=True, **gpuargs)
+
+    print('Training with %d image pairs' % len(train_dataset))
+    return train_loader
+
+def fetch_optimizer(args, model):
+    """ Create the optimizer and learning rate scheduler """
+    optimizer = optim.AdamW(model.parameters(), lr=args.lr, weight_decay=args.wdecay, eps=args.epsilon)
+
+    scheduler = optim.lr_scheduler.OneCycleLR(optimizer, args.lr, args.num_steps,
+        pct_start=0.2, cycle_momentum=False, anneal_strategy='linear', final_div_factor=0.05)
+
+    return optimizer, scheduler
+    
+
+class Logger:
+    def __init__(self, model, scheduler):
+        self.model = model
+        self.scheduler = scheduler
+        self.total_steps = 0
+        self.running_loss = {}
+
+    def _print_training_status(self):
+        metrics_data = [self.running_loss[k]/SUM_FREQ for k in sorted(self.running_loss.keys())]
+        training_str = "[{:6d}, {:10.7f}] ".format(self.total_steps+1, self.scheduler.get_lr()[0])
+        metrics_str = ("{:10.4f}, "*len(metrics_data)).format(*metrics_data)
+        
+        # print the training status
+        print(training_str + metrics_str)
+
+        for key in self.running_loss:
+            self.running_loss[key] = 0.0
+
+    def push(self, metrics):
+        self.total_steps += 1
+
+        for key in metrics:
+            if key not in self.running_loss:
+                self.running_loss[key] = 0.0
+
+            self.running_loss[key] += metrics[key]
+
+        if self.total_steps % SUM_FREQ == SUM_FREQ-1:
+            self._print_training_status()
+            self.running_loss = {}
+
+
+def train(args):
+
+    model = RAFT(args)
+    model = nn.DataParallel(model)
+    print("Parameter Count: %d" % count_parameters(model))
+
+    if args.restore_ckpt is not None:
+        model.load_state_dict(torch.load(args.restore_ckpt))
+
+    model.cuda()
+    model.train()
+    
+    if 'chairs' not in args.dataset:
+        model.module.freeze_bn()
+
+    train_loader = fetch_dataloader(args)
+    optimizer, scheduler = fetch_optimizer(args, model)
+
+    total_steps = 0
+    logger = Logger(model, scheduler)
+
+    should_keep_training = True
+    while should_keep_training:
+
+        for i_batch, data_blob in enumerate(train_loader):
+            image1, image2, flow, valid = [x.cuda() for x in data_blob]
+
+            optimizer.zero_grad()
+            flow_predictions = model(image1, image2, iters=args.iters)
+            
+            loss, metrics = sequence_loss(flow_predictions, flow, valid)
+            loss.backward()
+
+            torch.nn.utils.clip_grad_norm_(model.parameters(), args.clip)
+            optimizer.step()
+            scheduler.step()
+            total_steps += 1
+
+            logger.push(metrics)
+
+            if total_steps % VAL_FREQ == VAL_FREQ-1:
+                PATH = 'checkpoints/%d_%s.pth' % (total_steps+1, args.name)
+                torch.save(model.state_dict(), PATH)
+
+            if total_steps == args.num_steps:
+                should_keep_training = False
+                break
+
+
+    PATH = 'checkpoints/%s.pth' % args.name
+    torch.save(model.state_dict(), PATH)
+
+    return PATH
+
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--name', default='bla', help="name your experiment")
+    parser.add_argument('--dataset', help="which dataset to use for training") 
+    parser.add_argument('--restore_ckpt', help="restore checkpoint")
+    parser.add_argument('--small', action='store_true', help='use small model')
+
+    parser.add_argument('--lr', type=float, default=0.00002)
+    parser.add_argument('--num_steps', type=int, default=100000)
+    parser.add_argument('--batch_size', type=int, default=6)
+    parser.add_argument('--image_size', type=int, nargs='+', default=[384, 512])
+
+    parser.add_argument('--iters', type=int, default=12)
+    parser.add_argument('--wdecay', type=float, default=.00005)
+    parser.add_argument('--epsilon', type=float, default=1e-8)
+    parser.add_argument('--clip', type=float, default=1.0)
+    parser.add_argument('--dropout', type=float, default=0.0)
+    args = parser.parse_args()
+
+    torch.manual_seed(1234)
+    np.random.seed(1234)
+
+    if not os.path.isdir('checkpoints'):
+        os.mkdir('checkpoints')
+
+    # scale learning rate and batch size by number of GPUs
+    num_gpus = torch.cuda.device_count()
+    args.batch_size = args.batch_size * num_gpus
+    args.lr = args.lr * num_gpus
+
+    train(args)