Add new arxiv model and benchmark

README.md

@@ -117,16 +117,26 @@ docker run --rm -it --gpus all -v /dev/dri:/dev/dri -v $PWD:/host rife:latest in
 ## Evaluation
 **Our paper has not been officially published yet, and our method and experimental results are under improvement. Due to the incorrect data reference, the latency measurement of Sepconv and TOFlow in our arxiv paper needs to be modified.**
 
-Download [RIFE model](https://drive.google.com/file/d/1c1R7iF-ypN6USo-D2YH_ORtaH3tukSlo/view?usp=sharing) or [RIFE2F1.5C model](https://drive.google.com/file/d/1ve9w-cRWotdvvbU1KcgtsSm12l-JUkeT/view?usp=sharing) reported by our paper.
+Download [RIFE model](https://drive.google.com/file/d/1U2AGFY00hafsPmm94-6deeM-9feGN-qg/view?usp=sharing) reported by our paper.
+
+**UCF101**: Download [UCF101 dataset](https://liuziwei7.github.io/projects/VoxelFlow) at ./UCF101/ucf101_interp_ours/
 
 **Vimeo90K**: Download [Vimeo90K dataset](http://toflow.csail.mit.edu/) at ./vimeo_interp_test
 
 **MiddleBury**: Download [MiddleBury OTHER dataset](https://vision.middlebury.edu/flow/data/) at ./other-data and ./other-gt-interp
+
+**HD**: Download [HD dataset](https://github.com/baowenbo/MEMC-Net) at ./HD_dataset
 ```
+python3 benchmark/UCF101.py
+# (Final result: "Avg PSNR: 35.246 SSIM: 0.9691")
 python3 benchmark/Vimeo90K.py
-# (Final result: "Avg PSNR: 35.695 SSIM: 0.9788")
+# (Final result: "Avg PSNR: 35.506 SSIM: 0.9779")
 python3 benchmark/MiddelBury_Other.py
-# (Final result: "2.058")
+# (Final result: "1.962")
+python3 benchmark/HD.py
+# (Final result: "PSNR: 32.124")
+python3 benchmark/HD_multi.py
+# (Final result: "PSNR: 19.92(544*1280), 30.03(720p), 26.71(1080p)")
 ```
 
 ## Training and Reproduction

benchmark/HD_multi.py

@@ -0,0 +1,100 @@
+import os
+import sys
+sys.path.append('.')
+import cv2
+import math
+import torch
+import argparse
+import numpy as np
+from torch.nn import functional as F
+from pytorch_msssim import ssim_matlab
+from model.RIFE import Model
+from skimage.color import rgb2yuv, yuv2rgb
+from yuv_frame_io import YUV_Read,YUV_Write
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+
+model = Model()
+model.load_model('train_log')
+model.eval()
+model.device()
+
+name_list = [
+    ('HD_dataset/HD720p_GT/parkrun_1280x720_50.yuv', 720, 1280),
+    ('HD_dataset/HD720p_GT/shields_1280x720_60.yuv', 720, 1280),
+    ('HD_dataset/HD720p_GT/stockholm_1280x720_60.yuv', 720, 1280),
+    ('HD_dataset/HD1080p_GT/BlueSky.yuv', 1080, 1920),
+    ('HD_dataset/HD1080p_GT/Kimono1_1920x1080_24.yuv', 1080, 1920),
+    ('HD_dataset/HD1080p_GT/ParkScene_1920x1080_24.yuv', 1080, 1920),
+    ('HD_dataset/HD1080p_GT/sunflower_1080p25.yuv', 1080, 1920),
+    ('HD_dataset/HD544p_GT/Sintel_Alley2_1280x544.yuv', 544, 1280),
+    ('HD_dataset/HD544p_GT/Sintel_Market5_1280x544.yuv', 544, 1280),
+    ('HD_dataset/HD544p_GT/Sintel_Temple1_1280x544.yuv', 544, 1280),
+    ('HD_dataset/HD544p_GT/Sintel_Temple2_1280x544.yuv', 544, 1280),
+]
+def inference(I0, I1, pad, multi=3):
+    img = [I0, I1]
+    for i in range(multi):
+        res = [I0]
+        for j in range(len(img) - 1):
+            res.append(model.inference(img[j], img[j + 1]))
+            res.append(img[j + 1])
+        img = res
+    for i in range(len(img)):
+        img[i] = img[i][0][:, pad: -pad]
+    return img[1: -1]
+        
+tot = []
+for data in name_list:
+    psnr_list = []
+    name = data[0]
+    h = data[1]
+    w = data[2]
+    if 'yuv' in name:
+        Reader = YUV_Read(name, h, w, toRGB=True)
+    else:
+        Reader = cv2.VideoCapture(name)
+    _, lastframe = Reader.read()
+    # fourcc = cv2.VideoWriter_fourcc('m', 'p', '4', 'v')
+    # video = cv2.VideoWriter(name + '.mp4', fourcc, 30, (w, h))    
+    for index in range(0, 100, 8):
+        gt = []
+        if 'yuv' in name:
+            IMAGE1, success1 = Reader.read(index)
+            IMAGE2, success2 = Reader.read(index + 8)
+            if not success2:
+                break
+            for i in range(1, 8):
+                tmp, _ = Reader.read(index + i)
+                gt.append(tmp[:, :, ::-1].copy())
+            IMAGE1 = IMAGE1[:, :, ::-1].copy()
+            IMAGE2 = IMAGE2[:, :, ::-1].copy()
+        else:
+            print('Not Implement')
+        I0 = torch.from_numpy(np.transpose(IMAGE1, (2,0,1)).astype("float32") / 255.).cuda().unsqueeze(0)
+        I1 = torch.from_numpy(np.transpose(IMAGE2, (2,0,1)).astype("float32") / 255.).cuda().unsqueeze(0)
+        
+        if h == 720:
+            pad = 24
+        elif h == 1080:
+            pad = 4
+        else:
+            pad = 16
+        pader = torch.nn.ReplicationPad2d([0, 0, pad, pad])
+        I0 = pader(I0)
+        I1 = pader(I1)
+        with torch.no_grad():
+            pred = inference(I0, I1, pad)
+        for i in range(8 - 1):
+            out = (np.round(pred[i].detach().cpu().numpy().transpose(1, 2, 0) * 255)).astype('uint8')
+            if 'yuv' in name:
+                diff_rgb = 128.0 + rgb2yuv(gt[i] / 255.)[:, :, 0] * 255 - rgb2yuv(out / 255.)[:, :, 0] * 255
+                mse = np.mean((diff_rgb - 128.0) ** 2)
+                PIXEL_MAX = 255.0
+                psnr = 20 * math.log10(PIXEL_MAX / math.sqrt(mse))
+            else:
+                print('Not Implement')
+        psnr_list.append(psnr)
+    print(np.mean(psnr_list))
+    tot.append(np.mean(psnr_list))
+
+print('PSNR: {}(544*1280), {}(720p), {}(1080p)'.format(np.mean(tot[7:11]), np.mean(tot[:3]), np.mean(tot[3:7])))

model/IFNet.py

@@ -1,115 +1,76 @@
 import torch
-import numpy as np
 import torch.nn as nn
 import torch.nn.functional as F
 from model.warplayer import warp
 
-
-device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+def deconv(in_planes, out_planes, kernel_size=4, stride=2, padding=1):
+    return nn.Sequential(
+        torch.nn.ConvTranspose2d(in_channels=in_planes, out_channels=out_planes, kernel_size=4, stride=2, padding=1),
+        nn.PReLU(out_planes)
+    )
 
 def conv_wo_act(in_planes, out_planes, kernel_size=3, stride=1, padding=1, dilation=1):
     return nn.Sequential(
         nn.Conv2d(in_planes, out_planes, kernel_size=kernel_size, stride=stride,
-                  padding=padding, dilation=dilation, bias=False),
-        nn.BatchNorm2d(out_planes),
-    )
-
+                  padding=padding, dilation=dilation, bias=True),
+        )
 
 def conv(in_planes, out_planes, kernel_size=3, stride=1, padding=1, dilation=1):
     return nn.Sequential(
         nn.Conv2d(in_planes, out_planes, kernel_size=kernel_size, stride=stride,
-                  padding=padding, dilation=dilation, bias=False),
-        nn.BatchNorm2d(out_planes),
+                  padding=padding, dilation=dilation, bias=True),
         nn.PReLU(out_planes)
     )
 
-
-class ResBlock(nn.Module):
-    def __init__(self, in_planes, out_planes, stride=1):
-        super(ResBlock, self).__init__()
-        if in_planes == out_planes and stride == 1:
-            self.conv0 = nn.Identity()
-        else:
-            self.conv0 = nn.Conv2d(in_planes, out_planes,
-                                   3, stride, 1, bias=False)
-        self.conv1 = conv(in_planes, out_planes, 3, stride, 1)
-        self.conv2 = conv_wo_act(out_planes, out_planes, 3, 1, 1)
-        self.relu1 = nn.PReLU(1)
-        self.relu2 = nn.PReLU(out_planes)
-        self.fc1 = nn.Conv2d(out_planes, 16, kernel_size=1, bias=False)
-        self.fc2 = nn.Conv2d(16, out_planes, kernel_size=1, bias=False)
-
-    def forward(self, x):
-        y = self.conv0(x)
-        x = self.conv1(x)
-        x = self.conv2(x)
-        w = x.mean(3, True).mean(2, True)
-        w = self.relu1(self.fc1(w))
-        w = torch.sigmoid(self.fc2(w))
-        x = self.relu2(x * w + y)
-        return x
-
-
 class IFBlock(nn.Module):
     def __init__(self, in_planes, scale=1, c=64):
         super(IFBlock, self).__init__()
         self.scale = scale
-        self.conv0 = conv(in_planes, c, 3, 2, 1)
-        self.res0 = ResBlock(c, c)
-        self.res1 = ResBlock(c, c)
-        self.res2 = ResBlock(c, c)
-        self.res3 = ResBlock(c, c)
-        self.res4 = ResBlock(c, c)
-        self.res5 = ResBlock(c, c)
-        self.conv1 = nn.Conv2d(c, 8, 3, 1, 1)
-        self.up = nn.PixelShuffle(2)
+        self.conv0 = nn.Sequential(
+            conv(in_planes, c//2, 3, 2, 1),
+            conv(c//2, c, 3, 2, 1),
+            )
+        self.convblock = nn.Sequential(
+            conv(c, c),
+            conv(c, c),
+            conv(c, c),
+            conv(c, c),
+            conv(c, c),
+            conv(c, c),
+            conv(c, c),
+            conv(c, c),
+        )
+        self.conv1 = nn.ConvTranspose2d(c, 4, 4, 2, 1)
 
     def forward(self, x):
         if self.scale != 1:
-            x = F.interpolate(x, scale_factor=1. / self.scale, mode="bilinear",
-                              align_corners=False)
+            x = F.interpolate(x, scale_factor= 1. / self.scale, mode="bilinear", align_corners=False, recompute_scale_factor=False)
         x = self.conv0(x)
-        x = self.res0(x)
-        x = self.res1(x)
-        x = self.res2(x)
-        x = self.res3(x)
-        x = self.res4(x)
-        x = self.res5(x)
+        x = self.convblock(x) + x
         x = self.conv1(x)
-        flow = self.up(x)
+        flow = x
         if self.scale != 1:
-            flow = F.interpolate(flow, scale_factor=self.scale, mode="bilinear",
-                                 align_corners=False)
+            flow = F.interpolate(flow, scale_factor= self.scale, mode="bilinear", align_corners=False, recompute_scale_factor=False)
         return flow
-
-
+    
 class IFNet(nn.Module):
     def __init__(self):
         super(IFNet, self).__init__()
-        self.block0 = IFBlock(6, scale=4, c=192)
-        self.block1 = IFBlock(8, scale=2, c=128)
-        self.block2 = IFBlock(8, scale=1, c=64)
+        self.block0 = IFBlock(6, scale=4, c=240)
+        self.block1 = IFBlock(10, scale=2, c=150)
+        self.block2 = IFBlock(10, scale=1, c=90)
 
     def forward(self, x):
-        x = F.interpolate(x, scale_factor=0.5, mode="bilinear",
-                          align_corners=False)
         flow0 = self.block0(x)
         F1 = flow0
-        warped_img0 = warp(x[:, :3], F1)
-        warped_img1 = warp(x[:, 3:], -F1)
-        flow1 = self.block1(torch.cat((warped_img0, warped_img1, F1), 1))
+        F1_large = F.interpolate(F1, scale_factor=2.0, mode="bilinear", align_corners=False, recompute_scale_factor=False) * 2.0
+        warped_img0 = warp(x[:, :3], F1_large[:, :2])
+        warped_img1 = warp(x[:, 3:], F1_large[:, 2:4])
+        flow1 = self.block1(torch.cat((warped_img0, warped_img1, F1_large), 1))
         F2 = (flow0 + flow1)
-        warped_img0 = warp(x[:, :3], F2)
-        warped_img1 = warp(x[:, 3:], -F2)
-        flow2 = self.block2(torch.cat((warped_img0, warped_img1, F2), 1))
+        F2_large = F.interpolate(F2, scale_factor=2.0, mode="bilinear", align_corners=False, recompute_scale_factor=False) * 2.0
+        warped_img0 = warp(x[:, :3], F2_large[:, :2])
+        warped_img1 = warp(x[:, 3:], F2_large[:, 2:4])
+        flow2 = self.block2(torch.cat((warped_img0, warped_img1, F2_large), 1))
         F3 = (flow0 + flow1 + flow2)
         return F3, [F1, F2, F3]
-
-if __name__ == '__main__':
-    img0 = torch.zeros(3, 3, 256, 256).float().to(device)
-    img1 = torch.tensor(np.random.normal(
-        0, 1, (3, 3, 256, 256))).float().to(device)
-    imgs = torch.cat((img0, img1), 1)
-    flownet = IFNet()
-    flow, _ = flownet(imgs)
-    print(flow.shape)

model/IFNet15C.py

@@ -0,0 +1,76 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from model.warplayer import warp
+
+def deconv(in_planes, out_planes, kernel_size=4, stride=2, padding=1):
+    return nn.Sequential(
+        torch.nn.ConvTranspose2d(in_channels=in_planes, out_channels=out_planes, kernel_size=4, stride=2, padding=1),
+        nn.PReLU(out_planes)
+    )
+
+def conv_wo_act(in_planes, out_planes, kernel_size=3, stride=1, padding=1, dilation=1):
+    return nn.Sequential(
+        nn.Conv2d(in_planes, out_planes, kernel_size=kernel_size, stride=stride,
+                  padding=padding, dilation=dilation, bias=True),
+        )
+
+def conv(in_planes, out_planes, kernel_size=3, stride=1, padding=1, dilation=1):
+    return nn.Sequential(
+        nn.Conv2d(in_planes, out_planes, kernel_size=kernel_size, stride=stride,
+                  padding=padding, dilation=dilation, bias=True),
+        nn.PReLU(out_planes)
+    )
+
+class IFBlock(nn.Module):
+    def __init__(self, in_planes, scale=1, c=64):
+        super(IFBlock, self).__init__()
+        self.scale = scale
+        self.conv0 = nn.Sequential(
+            conv(in_planes, c//2, 3, 2, 1),
+            conv(c//2, c, 3, 2, 1),
+            )
+        self.convblock = nn.Sequential(
+            conv(c, c),
+            conv(c, c),
+            conv(c, c),
+            conv(c, c),
+            conv(c, c),
+            conv(c, c),
+            conv(c, c),
+            conv(c, c),
+        )
+        self.conv1 = nn.ConvTranspose2d(c, 4, 4, 2, 1)
+
+    def forward(self, x):
+        if self.scale != 1:
+            x = F.interpolate(x, scale_factor= 1. / self.scale, mode="bilinear", align_corners=False, recompute_scale_factor=False)
+        x = self.conv0(x)
+        x = self.convblock(x) + x
+        x = self.conv1(x)
+        flow = x
+        if self.scale != 1:
+            flow = F.interpolate(flow, scale_factor= self.scale, mode="bilinear", align_corners=False, recompute_scale_factor=False)
+        return flow
+    
+class IFNet(nn.Module):
+    def __init__(self):
+        super(IFNet, self).__init__()
+        self.block0 = IFBlock(6, scale=4, c=320)
+        self.block1 = IFBlock(10, scale=2, c=225)
+        self.block2 = IFBlock(10, scale=1, c=135)
+
+    def forward(self, x):
+        flow0 = self.block0(x)
+        F1 = flow0
+        F1_large = F.interpolate(F1, scale_factor=2.0, mode="bilinear", align_corners=False, recompute_scale_factor=False) * 2.0
+        warped_img0 = warp(x[:, :3], F1_large[:, :2])
+        warped_img1 = warp(x[:, 3:], F1_large[:, 2:4])
+        flow1 = self.block1(torch.cat((warped_img0, warped_img1, F1_large), 1))
+        F2 = (flow0 + flow1)
+        F2_large = F.interpolate(F2, scale_factor=2.0, mode="bilinear", align_corners=False, recompute_scale_factor=False) * 2.0
+        warped_img0 = warp(x[:, :3], F2_large[:, :2])
+        warped_img1 = warp(x[:, 3:], F2_large[:, 2:4])
+        flow2 = self.block2(torch.cat((warped_img0, warped_img1, F2_large), 1))
+        F3 = (flow0 + flow1 + flow2)
+        return F3, [F1, F2, F3]

model/IFNet2F.py

@@ -1,115 +1,75 @@
 import torch
-import numpy as np
 import torch.nn as nn
 import torch.nn.functional as F
 from model.warplayer import warp
 
-
-device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+def deconv(in_planes, out_planes, kernel_size=4, stride=2, padding=1):
+    return nn.Sequential(
+        torch.nn.ConvTranspose2d(in_channels=in_planes, out_channels=out_planes, kernel_size=4, stride=2, padding=1),
+        nn.PReLU(out_planes)
+    )
 
 def conv_wo_act(in_planes, out_planes, kernel_size=3, stride=1, padding=1, dilation=1):
     return nn.Sequential(
         nn.Conv2d(in_planes, out_planes, kernel_size=kernel_size, stride=stride,
-                  padding=padding, dilation=dilation, bias=False),
-        nn.BatchNorm2d(out_planes),
-    )
-
+                  padding=padding, dilation=dilation, bias=True),
+        )
 
 def conv(in_planes, out_planes, kernel_size=3, stride=1, padding=1, dilation=1):
     return nn.Sequential(
         nn.Conv2d(in_planes, out_planes, kernel_size=kernel_size, stride=stride,
-                  padding=padding, dilation=dilation, bias=False),
-        nn.BatchNorm2d(out_planes),
+                  padding=padding, dilation=dilation, bias=True),
         nn.PReLU(out_planes)
     )
 
-
-class ResBlock(nn.Module):
-    def __init__(self, in_planes, out_planes, stride=1):
-        super(ResBlock, self).__init__()
-        if in_planes == out_planes and stride == 1:
-            self.conv0 = nn.Identity()
-        else:
-            self.conv0 = nn.Conv2d(in_planes, out_planes,
-                                   3, stride, 1, bias=False)
-        self.conv1 = conv(in_planes, out_planes, 3, stride, 1)
-        self.conv2 = conv_wo_act(out_planes, out_planes, 3, 1, 1)
-        self.relu1 = nn.PReLU(1)
-        self.relu2 = nn.PReLU(out_planes)
-        self.fc1 = nn.Conv2d(out_planes, 16, kernel_size=1, bias=False)
-        self.fc2 = nn.Conv2d(16, out_planes, kernel_size=1, bias=False)
-
-    def forward(self, x):
-        y = self.conv0(x)
-        x = self.conv1(x)
-        x = self.conv2(x)
-        w = x.mean(3, True).mean(2, True)
-        w = self.relu1(self.fc1(w))
-        w = torch.sigmoid(self.fc2(w))
-        x = self.relu2(x * w + y)
-        return x
-
-
 class IFBlock(nn.Module):
     def __init__(self, in_planes, scale=1, c=64):
         super(IFBlock, self).__init__()
         self.scale = scale
-        self.conv0 = conv(in_planes, c, 3, 1, 1)
-        self.res0 = ResBlock(c, c)
-        self.res1 = ResBlock(c, c)
-        self.res2 = ResBlock(c, c)
-        self.res3 = ResBlock(c, c)
-        self.res4 = ResBlock(c, c)
-        self.res5 = ResBlock(c, c)
-        self.conv1 = nn.Conv2d(c, 2, 3, 1, 1)
-        self.up = nn.PixelShuffle(2)
+        self.conv0 = nn.Sequential(
+            conv(in_planes, c, 3, 2, 1),
+            )
+        self.convblock = nn.Sequential(
+            conv(c, c),
+            conv(c, c),
+            conv(c, c),
+            conv(c, c),
+            conv(c, c),
+            conv(c, c),
+            conv(c, c),
+            conv(c, c),
+        )
+        self.conv1 = nn.Conv2d(c, 4, 3, 1, 1)
 
     def forward(self, x):
         if self.scale != 1:
-            x = F.interpolate(x, scale_factor=1. / self.scale, mode="bilinear",
-                              align_corners=False)
+            x = F.interpolate(x, scale_factor= 1. / self.scale, mode="bilinear", align_corners=False, recompute_scale_factor=False)
         x = self.conv0(x)
-        x = self.res0(x)
-        x = self.res1(x)
-        x = self.res2(x)
-        x = self.res3(x)
-        x = self.res4(x)
-        x = self.res5(x)
+        x = self.convblock(x) + x
         x = self.conv1(x)
-        flow = x # self.up(x)
+        flow = x
         if self.scale != 1:
-            flow = F.interpolate(flow, scale_factor=self.scale, mode="bilinear",
-                                 align_corners=False)
+            flow = F.interpolate(flow, scale_factor= self.scale, mode="bilinear", align_corners=False, recompute_scale_factor=False)
         return flow
-
-
+    
 class IFNet(nn.Module):
     def __init__(self):
         super(IFNet, self).__init__()
-        self.block0 = IFBlock(6, scale=4, c=192)
-        self.block1 = IFBlock(8, scale=2, c=128)
-        self.block2 = IFBlock(8, scale=1, c=64)
+        self.block0 = IFBlock(6, scale=4, c=240)
+        self.block1 = IFBlock(10, scale=2, c=150)
+        self.block2 = IFBlock(10, scale=1, c=90)
 
     def forward(self, x):
-        x = F.interpolate(x, scale_factor=0.5, mode="bilinear",
-                          align_corners=False)
         flow0 = self.block0(x)
         F1 = flow0
-        warped_img0 = warp(x[:, :3], F1)
-        warped_img1 = warp(x[:, 3:], -F1)
-        flow1 = self.block1(torch.cat((warped_img0, warped_img1, F1), 1))
+        F1_large = F.interpolate(F1, scale_factor=2.0, mode="bilinear", align_corners=False, recompute_scale_factor=False) * 2.0
+        warped_img0 = warp(x[:, :3], F1_large[:, :2])
+        warped_img1 = warp(x[:, 3:], F1_large[:, 2:4])
+        flow1 = self.block1(torch.cat((warped_img0, warped_img1, F1_large), 1))
         F2 = (flow0 + flow1)
-        warped_img0 = warp(x[:, :3], F2)
-        warped_img1 = warp(x[:, 3:], -F2)
-        flow2 = self.block2(torch.cat((warped_img0, warped_img1, F2), 1))
+        F2_large = F.interpolate(F2, scale_factor=2.0, mode="bilinear", align_corners=False, recompute_scale_factor=False) * 2.0
+        warped_img0 = warp(x[:, :3], F2_large[:, :2])
+        warped_img1 = warp(x[:, 3:], F2_large[:, 2:4])
+        flow2 = self.block2(torch.cat((warped_img0, warped_img1, F2_large), 1))
         F3 = (flow0 + flow1 + flow2)
         return F3, [F1, F2, F3]
-
-if __name__ == '__main__':
-    img0 = torch.zeros(3, 3, 256, 256).float().to(device)
-    img1 = torch.tensor(np.random.normal(
-        0, 1, (3, 3, 256, 256))).float().to(device)
-    imgs = torch.cat((img0, img1), 1)
-    flownet = IFNet()
-    flow, _ = flownet(imgs)
-    print(flow.shape)

model/IFNet2F15C.py

@@ -1,115 +1,75 @@
 import torch
-import numpy as np
 import torch.nn as nn
 import torch.nn.functional as F
 from model.warplayer import warp
 
-
-device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+def deconv(in_planes, out_planes, kernel_size=4, stride=2, padding=1):
+    return nn.Sequential(
+        torch.nn.ConvTranspose2d(in_channels=in_planes, out_channels=out_planes, kernel_size=4, stride=2, padding=1),
+        nn.PReLU(out_planes)
+    )
 
 def conv_wo_act(in_planes, out_planes, kernel_size=3, stride=1, padding=1, dilation=1):
     return nn.Sequential(
         nn.Conv2d(in_planes, out_planes, kernel_size=kernel_size, stride=stride,
-                  padding=padding, dilation=dilation, bias=False),
-        nn.BatchNorm2d(out_planes),
-    )
-
+                  padding=padding, dilation=dilation, bias=True),
+        )
 
 def conv(in_planes, out_planes, kernel_size=3, stride=1, padding=1, dilation=1):
     return nn.Sequential(
         nn.Conv2d(in_planes, out_planes, kernel_size=kernel_size, stride=stride,
-                  padding=padding, dilation=dilation, bias=False),
-        nn.BatchNorm2d(out_planes),
+                  padding=padding, dilation=dilation, bias=True),
         nn.PReLU(out_planes)
     )
 
-
-class ResBlock(nn.Module):
-    def __init__(self, in_planes, out_planes, stride=1):
-        super(ResBlock, self).__init__()
-        if in_planes == out_planes and stride == 1:
-            self.conv0 = nn.Identity()
-        else:
-            self.conv0 = nn.Conv2d(in_planes, out_planes,
-                                   3, stride, 1, bias=False)
-        self.conv1 = conv(in_planes, out_planes, 3, stride, 1)
-        self.conv2 = conv_wo_act(out_planes, out_planes, 3, 1, 1)
-        self.relu1 = nn.PReLU(1)
-        self.relu2 = nn.PReLU(out_planes)
-        self.fc1 = nn.Conv2d(out_planes, 16, kernel_size=1, bias=False)
-        self.fc2 = nn.Conv2d(16, out_planes, kernel_size=1, bias=False)
-
-    def forward(self, x):
-        y = self.conv0(x)
-        x = self.conv1(x)
-        x = self.conv2(x)
-        w = x.mean(3, True).mean(2, True)
-        w = self.relu1(self.fc1(w))
-        w = torch.sigmoid(self.fc2(w))
-        x = self.relu2(x * w + y)
-        return x
-
-
 class IFBlock(nn.Module):
     def __init__(self, in_planes, scale=1, c=64):
         super(IFBlock, self).__init__()
         self.scale = scale
-        self.conv0 = conv(in_planes, c, 3, 1, 1)
-        self.res0 = ResBlock(c, c)
-        self.res1 = ResBlock(c, c)
-        self.res2 = ResBlock(c, c)
-        self.res3 = ResBlock(c, c)
-        self.res4 = ResBlock(c, c)
-        self.res5 = ResBlock(c, c)
-        self.conv1 = nn.Conv2d(c, 2, 3, 1, 1)
-        self.up = nn.PixelShuffle(2)
+        self.conv0 = nn.Sequential(
+            conv(in_planes, c, 3, 2, 1),
+            )
+        self.convblock = nn.Sequential(
+            conv(c, c),
+            conv(c, c),
+            conv(c, c),
+            conv(c, c),
+            conv(c, c),
+            conv(c, c),
+            conv(c, c),
+            conv(c, c),
+        )
+        self.conv1 = nn.Conv2d(c, 4, 3, 1, 1)
 
     def forward(self, x):
         if self.scale != 1:
-            x = F.interpolate(x, scale_factor=1. / self.scale, mode="bilinear",
-                              align_corners=False)
+            x = F.interpolate(x, scale_factor= 1. / self.scale, mode="bilinear", align_corners=False, recompute_scale_factor=False)
         x = self.conv0(x)
-        x = self.res0(x)
-        x = self.res1(x)
-        x = self.res2(x)
-        x = self.res3(x)
-        x = self.res4(x)
-        x = self.res5(x)
+        x = self.convblock(x) + x
         x = self.conv1(x)
-        flow = x # self.up(x)
+        flow = x
         if self.scale != 1:
-            flow = F.interpolate(flow, scale_factor=self.scale, mode="bilinear",
-                                 align_corners=False)
+            flow = F.interpolate(flow, scale_factor= self.scale, mode="bilinear", align_corners=False, recompute_scale_factor=False)
         return flow
-
-
+    
 class IFNet(nn.Module):
     def __init__(self):
         super(IFNet, self).__init__()
-        self.block0 = IFBlock(6, scale=4, c=288)
-        self.block1 = IFBlock(8, scale=2, c=192)
-        self.block2 = IFBlock(8, scale=1, c=96)
+        self.block0 = IFBlock(6, scale=4, c=360)
+        self.block1 = IFBlock(10, scale=2, c=225)
+        self.block2 = IFBlock(10, scale=1, c=135)
 
     def forward(self, x):
-        x = F.interpolate(x, scale_factor=0.5, mode="bilinear",
-                          align_corners=False)
         flow0 = self.block0(x)
         F1 = flow0
-        warped_img0 = warp(x[:, :3], F1)
-        warped_img1 = warp(x[:, 3:], -F1)
-        flow1 = self.block1(torch.cat((warped_img0, warped_img1, F1), 1))
+        F1_large = F.interpolate(F1, scale_factor=2.0, mode="bilinear", align_corners=False, recompute_scale_factor=False) * 2.0
+        warped_img0 = warp(x[:, :3], F1_large[:, :2])
+        warped_img1 = warp(x[:, 3:], F1_large[:, 2:4])
+        flow1 = self.block1(torch.cat((warped_img0, warped_img1, F1_large), 1))
         F2 = (flow0 + flow1)
-        warped_img0 = warp(x[:, :3], F2)
-        warped_img1 = warp(x[:, 3:], -F2)
-        flow2 = self.block2(torch.cat((warped_img0, warped_img1, F2), 1))
+        F2_large = F.interpolate(F2, scale_factor=2.0, mode="bilinear", align_corners=False, recompute_scale_factor=False) * 2.0
+        warped_img0 = warp(x[:, :3], F2_large[:, :2])
+        warped_img1 = warp(x[:, 3:], F2_large[:, 2:4])
+        flow2 = self.block2(torch.cat((warped_img0, warped_img1, F2_large), 1))
         F3 = (flow0 + flow1 + flow2)
         return F3, [F1, F2, F3]
-
-if __name__ == '__main__':
-    img0 = torch.zeros(3, 3, 256, 256).float().to(device)
-    img1 = torch.tensor(np.random.normal(
-        0, 1, (3, 3, 256, 256))).float().to(device)
-    imgs = torch.cat((img0, img1), 1)
-    flownet = IFNet()
-    flow, _ = flownet(imgs)
-    print(flow.shape)

model/RIFE.py

@@ -34,29 +34,15 @@ def conv_woact(in_planes, out_planes, kernel_size=3, stride=1, padding=1, dilati
                   padding=padding, dilation=dilation, bias=True),
     )
 
-class ResBlock(nn.Module):
+class Conv2(nn.Module):
     def __init__(self, in_planes, out_planes, stride=2):
-        super(ResBlock, self).__init__()
-        if in_planes == out_planes and stride == 1:
-            self.conv0 = nn.Identity()
-        else:
-            self.conv0 = nn.Conv2d(in_planes, out_planes,
-                                   3, stride, 1, bias=False)
+        super(Conv2, self).__init__()
         self.conv1 = conv(in_planes, out_planes, 3, stride, 1)
-        self.conv2 = conv_woact(out_planes, out_planes, 3, 1, 1)
-        self.relu1 = nn.PReLU(1)
-        self.relu2 = nn.PReLU(out_planes)
-        self.fc1 = nn.Conv2d(out_planes, 16, kernel_size=1, bias=False)
-        self.fc2 = nn.Conv2d(16, out_planes, kernel_size=1, bias=False)
-
+        self.conv2 = conv(out_planes, out_planes, 3, 1, 1)
+        
     def forward(self, x):
-        y = self.conv0(x)
         x = self.conv1(x)
         x = self.conv2(x)
-        w = x.mean(3, True).mean(2, True)
-        w = self.relu1(self.fc1(w))
-        w = torch.sigmoid(self.fc2(w))
-        x = self.relu2(x * w + y)
         return x
 
 c = 16
@@ -64,36 +50,32 @@ c = 16
 class ContextNet(nn.Module):
     def __init__(self):
         super(ContextNet, self).__init__()
-        self.conv1 = ResBlock(3, c)
-        self.conv2 = ResBlock(c, 2*c)
-        self.conv3 = ResBlock(2*c, 4*c)
-        self.conv4 = ResBlock(4*c, 8*c)
+        self.conv1 = Conv2(3, c)
+        self.conv2 = Conv2(c, 2*c)
+        self.conv3 = Conv2(2*c, 4*c)
+        self.conv4 = Conv2(4*c, 8*c)
 
     def forward(self, x, flow):
         x = self.conv1(x)
         f1 = warp(x, flow)
         x = self.conv2(x)
-        flow = F.interpolate(flow, scale_factor=0.5, mode="bilinear",
-                             align_corners=False) * 0.5
+        flow = F.interpolate(flow, scale_factor=0.5, mode="bilinear", align_corners=False, recompute_scale_factor=False) * 0.5
         f2 = warp(x, flow)
         x = self.conv3(x)
-        flow = F.interpolate(flow, scale_factor=0.5, mode="bilinear",
-                             align_corners=False) * 0.5
+        flow = F.interpolate(flow, scale_factor=0.5, mode="bilinear", align_corners=False, recompute_scale_factor=False) * 0.5
         f3 = warp(x, flow)
         x = self.conv4(x)
-        flow = F.interpolate(flow, scale_factor=0.5, mode="bilinear",
-                             align_corners=False) * 0.5
+        flow = F.interpolate(flow, scale_factor=0.5, mode="bilinear", align_corners=False, recompute_scale_factor=False) * 0.5
         f4 = warp(x, flow)
         return [f1, f2, f3, f4]
 
-
 class FusionNet(nn.Module):
     def __init__(self):
         super(FusionNet, self).__init__()
-        self.down0 = ResBlock(8, 2*c)
-        self.down1 = ResBlock(4*c, 4*c)
-        self.down2 = ResBlock(8*c, 8*c)
-        self.down3 = ResBlock(16*c, 16*c)
+        self.down0 = Conv2(12, 2*c)
+        self.down1 = Conv2(4*c, 4*c)
+        self.down2 = Conv2(8*c, 8*c)
+        self.down3 = Conv2(16*c, 16*c)
         self.up0 = deconv(32*c, 8*c)
         self.up1 = deconv(16*c, 4*c)
         self.up2 = deconv(8*c, 2*c)
@@ -101,14 +83,14 @@ class FusionNet(nn.Module):
         self.conv = nn.Conv2d(c, 4, 3, 1, 1)
 
     def forward(self, img0, img1, flow, c0, c1, flow_gt):
-        warped_img0 = warp(img0, flow)
-        warped_img1 = warp(img1, -flow)
+        warped_img0 = warp(img0, flow[:, :2])
+        warped_img1 = warp(img1, flow[:, 2:4])
         if flow_gt == None:
             warped_img0_gt, warped_img1_gt = None, None
         else:
             warped_img0_gt = warp(img0, flow_gt[:, :2])
             warped_img1_gt = warp(img1, flow_gt[:, 2:4])
-        s0 = self.down0(torch.cat((warped_img0, warped_img1, flow), 1))
+        s0 = self.down0(torch.cat((img0, img1, warped_img0, warped_img1), 1))
         s1 = self.down1(torch.cat((s0, c0[0], c1[0]), 1))
         s2 = self.down2(torch.cat((s1, c0[1], c1[1]), 1))
         s3 = self.down3(torch.cat((s2, c0[2], c1[2]), 1))
@@ -119,7 +101,6 @@ class FusionNet(nn.Module):
         x = self.conv(x)
         return x, warped_img0, warped_img1, warped_img0_gt, warped_img1_gt
 
-
 class Model:
     def __init__(self, local_rank=-1):
         self.flownet = IFNet()
@@ -158,7 +139,7 @@ class Model:
         self.contextnet.to(device)
         self.fusionnet.to(device)
 
-    def load_model(self, path, rank):
+    def load_model(self, path, rank=-1):
         def convert(param):
             if rank == -1:
                 return {
@@ -185,8 +166,8 @@ class Model:
     def predict(self, imgs, flow, training=True, flow_gt=None):
         img0 = imgs[:, :3]
         img1 = imgs[:, 3:]
-        c0 = self.contextnet(img0, flow)
-        c1 = self.contextnet(img1, -flow)
+        c0 = self.contextnet(img0, flow[:, :2])
+        c1 = self.contextnet(img1, flow[:, 2:4])
         flow = F.interpolate(flow, scale_factor=2.0, mode="bilinear",
                              align_corners=False) * 2.0
         refine_output, warped_img0, warped_img1, warped_img0_gt, warped_img1_gt = self.fusionnet(
@@ -203,7 +184,7 @@ class Model:
 
     def inference(self, img0, img1):
         imgs = torch.cat((img0, img1), 1)
-        flow, _ = self.flownet(imgs)
+        flow, _ = self.flownet(torch.cat((img0, img1), 1))
         return self.predict(imgs, flow, training=False)
 
     def update(self, imgs, gt, learning_rate=0, mul=1, training=True, flow_gt=None):
@@ -228,8 +209,8 @@ class Model:
                                          align_corners=False) * 0.5).detach()
             loss_cons = 0
             for i in range(3):
-                loss_cons += self.epe(flow_list[i], flow_gt[:, :2], 1)
-                loss_cons += self.epe(-flow_list[i], flow_gt[:, 2:4], 1)
+                loss_cons += self.epe(flow_list[i][:, :2], flow_gt[:, :2], 1)
+                loss_cons += self.epe(flow_list[i][:, 2:4], flow_gt[:, 2:4], 1)
             loss_cons = loss_cons.mean() * 0.01
         else:
             loss_cons = torch.tensor([0])

model/RIFE15C.py

@@ -0,0 +1,235 @@
+import torch
+import torch.nn as nn
+import numpy as np
+from torch.optim import AdamW
+import torch.optim as optim
+import itertools
+from model.warplayer import warp
+from torch.nn.parallel import DistributedDataParallel as DDP
+from model.IFNet import *
+import torch.nn.functional as F
+from model.loss import *
+
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+
+
+def conv(in_planes, out_planes, kernel_size=3, stride=1, padding=1, dilation=1):
+    return nn.Sequential(
+        nn.Conv2d(in_planes, out_planes, kernel_size=kernel_size, stride=stride,
+                  padding=padding, dilation=dilation, bias=True),
+        nn.PReLU(out_planes)
+    )
+
+
+def deconv(in_planes, out_planes, kernel_size=4, stride=2, padding=1):
+    return nn.Sequential(
+        torch.nn.ConvTranspose2d(in_channels=in_planes, out_channels=out_planes,
+                                 kernel_size=4, stride=2, padding=1, bias=True),
+        nn.PReLU(out_planes)
+    )
+
+def conv_woact(in_planes, out_planes, kernel_size=3, stride=1, padding=1, dilation=1):
+    return nn.Sequential(
+        nn.Conv2d(in_planes, out_planes, kernel_size=kernel_size, stride=stride,
+                  padding=padding, dilation=dilation, bias=True),
+    )
+
+class Conv2(nn.Module):
+    def __init__(self, in_planes, out_planes, stride=2):
+        super(Conv2, self).__init__()
+        self.conv1 = conv(in_planes, out_planes, 3, stride, 1)
+        self.conv2 = conv(out_planes, out_planes, 3, 1, 1)
+        
+    def forward(self, x):
+        x = self.conv1(x)
+        x = self.conv2(x)
+        return x
+
+c = 24
+
+class ContextNet(nn.Module):
+    def __init__(self):
+        super(ContextNet, self).__init__()
+        self.conv1 = Conv2(3, c)
+        self.conv2 = Conv2(c, 2*c)
+        self.conv3 = Conv2(2*c, 4*c)
+        self.conv4 = Conv2(4*c, 8*c)
+
+    def forward(self, x, flow):
+        x = self.conv1(x)
+        f1 = warp(x, flow)
+        x = self.conv2(x)
+        flow = F.interpolate(flow, scale_factor=0.5, mode="bilinear", align_corners=False, recompute_scale_factor=False) * 0.5
+        f2 = warp(x, flow)
+        x = self.conv3(x)
+        flow = F.interpolate(flow, scale_factor=0.5, mode="bilinear", align_corners=False, recompute_scale_factor=False) * 0.5
+        f3 = warp(x, flow)
+        x = self.conv4(x)
+        flow = F.interpolate(flow, scale_factor=0.5, mode="bilinear", align_corners=False, recompute_scale_factor=False) * 0.5
+        f4 = warp(x, flow)
+        return [f1, f2, f3, f4]
+
+class FusionNet(nn.Module):
+    def __init__(self):
+        super(FusionNet, self).__init__()
+        self.down0 = Conv2(12, 2*c)
+        self.down1 = Conv2(4*c, 4*c)
+        self.down2 = Conv2(8*c, 8*c)
+        self.down3 = Conv2(16*c, 16*c)
+        self.up0 = deconv(32*c, 8*c)
+        self.up1 = deconv(16*c, 4*c)
+        self.up2 = deconv(8*c, 2*c)
+        self.up3 = deconv(4*c, c)
+        self.conv = nn.Conv2d(c, 4, 3, 1, 1)
+
+    def forward(self, img0, img1, flow, c0, c1, flow_gt):
+        warped_img0 = warp(img0, flow[:, :2])
+        warped_img1 = warp(img1, flow[:, 2:4])
+        if flow_gt == None:
+            warped_img0_gt, warped_img1_gt = None, None
+        else:
+            warped_img0_gt = warp(img0, flow_gt[:, :2])
+            warped_img1_gt = warp(img1, flow_gt[:, 2:4])
+        s0 = self.down0(torch.cat((img0, img1, warped_img0, warped_img1), 1))
+        s1 = self.down1(torch.cat((s0, c0[0], c1[0]), 1))
+        s2 = self.down2(torch.cat((s1, c0[1], c1[1]), 1))
+        s3 = self.down3(torch.cat((s2, c0[2], c1[2]), 1))
+        x = self.up0(torch.cat((s3, c0[3], c1[3]), 1))
+        x = self.up1(torch.cat((x, s2), 1))
+        x = self.up2(torch.cat((x, s1), 1))
+        x = self.up3(torch.cat((x, s0), 1))
+        x = self.conv(x)
+        return x, warped_img0, warped_img1, warped_img0_gt, warped_img1_gt
+
+class Model:
+    def __init__(self, local_rank=-1):
+        self.flownet = IFNet()
+        self.contextnet = ContextNet()
+        self.fusionnet = FusionNet()
+        self.device()
+        self.optimG = AdamW(itertools.chain(
+            self.flownet.parameters(),
+            self.contextnet.parameters(),
+            self.fusionnet.parameters()), lr=1e-6, weight_decay=1e-5)
+        self.schedulerG = optim.lr_scheduler.CyclicLR(
+            self.optimG, base_lr=1e-6, max_lr=1e-3, step_size_up=8000, cycle_momentum=False)
+        self.epe = EPE()
+        self.ter = Ternary()
+        self.sobel = SOBEL()
+        if local_rank != -1:
+            self.flownet = DDP(self.flownet, device_ids=[
+                               local_rank], output_device=local_rank)
+            self.contextnet = DDP(self.contextnet, device_ids=[
+                                  local_rank], output_device=local_rank)
+            self.fusionnet = DDP(self.fusionnet, device_ids=[
+                                 local_rank], output_device=local_rank)
+
+    def train(self):
+        self.flownet.train()
+        self.contextnet.train()
+        self.fusionnet.train()
+
+    def eval(self):
+        self.flownet.eval()
+        self.contextnet.eval()
+        self.fusionnet.eval()
+
+    def device(self):
+        self.flownet.to(device)
+        self.contextnet.to(device)
+        self.fusionnet.to(device)
+
+    def load_model(self, path, rank=-1):
+        def convert(param):
+            if rank == -1:
+                return {
+                    k.replace("module.", ""): v
+                    for k, v in param.items()
+                    if "module." in k
+                }
+            else:
+                return param
+        if rank <= 0:
+            self.flownet.load_state_dict(
+                convert(torch.load('{}/flownet.pkl'.format(path), map_location=device)))
+            self.contextnet.load_state_dict(
+                convert(torch.load('{}/contextnet.pkl'.format(path), map_location=device)))
+            self.fusionnet.load_state_dict(
+                convert(torch.load('{}/unet.pkl'.format(path), map_location=device)))
+
+    def save_model(self, path, rank):
+        if rank == 0:
+            torch.save(self.flownet.state_dict(), '{}/flownet.pkl'.format(path))
+            torch.save(self.contextnet.state_dict(), '{}/contextnet.pkl'.format(path))
+            torch.save(self.fusionnet.state_dict(), '{}/unet.pkl'.format(path))
+
+    def predict(self, imgs, flow, training=True, flow_gt=None):
+        img0 = imgs[:, :3]
+        img1 = imgs[:, 3:]
+        c0 = self.contextnet(img0, flow[:, :2])
+        c1 = self.contextnet(img1, flow[:, 2:4])
+        flow = F.interpolate(flow, scale_factor=2.0, mode="bilinear",
+                             align_corners=False) * 2.0
+        refine_output, warped_img0, warped_img1, warped_img0_gt, warped_img1_gt = self.fusionnet(
+            img0, img1, flow, c0, c1, flow_gt)
+        res = torch.sigmoid(refine_output[:, :3]) * 2 - 1
+        mask = torch.sigmoid(refine_output[:, 3:4])
+        merged_img = warped_img0 * mask + warped_img1 * (1 - mask)
+        pred = merged_img + res
+        pred = torch.clamp(pred, 0, 1)
+        if training:
+            return pred, mask, merged_img, warped_img0, warped_img1, warped_img0_gt, warped_img1_gt
+        else:
+            return pred
+
+    def inference(self, img0, img1):
+        imgs = torch.cat((img0, img1), 1)
+        flow, _ = self.flownet(imgs)
+        return self.predict(imgs, flow, training=False)
+
+    def update(self, imgs, gt, learning_rate=0, mul=1, training=True, flow_gt=None):
+        for param_group in self.optimG.param_groups:
+            param_group['lr'] = learning_rate
+        if training:
+            self.train()
+        else:
+            self.eval()
+        flow, flow_list = self.flownet(imgs)
+        pred, mask, merged_img, warped_img0, warped_img1, warped_img0_gt, warped_img1_gt = self.predict(
+            imgs, flow, flow_gt=flow_gt)
+        loss_ter = self.ter(pred, gt).mean()
+        if training:
+            with torch.no_grad():
+                loss_flow = torch.abs(warped_img0_gt - gt).mean()
+                loss_mask = torch.abs(
+                    merged_img - gt).sum(1, True).float().detach()
+                loss_mask = F.interpolate(loss_mask, scale_factor=0.5, mode="bilinear",
+                                          align_corners=False).detach()
+                flow_gt = (F.interpolate(flow_gt, scale_factor=0.5, mode="bilinear",
+                                         align_corners=False) * 0.5).detach()
+            loss_cons = 0
+            for i in range(3):
+                loss_cons += self.epe(flow_list[i][:, :2], flow_gt[:, :2], 1)
+                loss_cons += self.epe(flow_list[i][:, 2:4], flow_gt[:, 2:4], 1)
+            loss_cons = loss_cons.mean() * 0.01
+        else:
+            loss_cons = torch.tensor([0])
+            loss_flow = torch.abs(warped_img0 - gt).mean()
+            loss_mask = 1
+        loss_l1 = (((pred - gt) ** 2 + 1e-6) ** 0.5).mean()
+        if training:
+            self.optimG.zero_grad()
+            loss_G = loss_l1 + loss_cons + loss_ter
+            loss_G.backward()
+            self.optimG.step()
+        return pred, merged_img, flow, loss_l1, loss_flow, loss_cons, loss_ter, loss_mask
+
+
+if __name__ == '__main__':
+    img0 = torch.zeros(3, 3, 256, 256).float().to(device)
+    img1 = torch.tensor(np.random.normal(
+        0, 1, (3, 3, 256, 256))).float().to(device)
+    imgs = torch.cat((img0, img1), 1)
+    model = Model()
+    model.eval()
+    print(model.inference(imgs).shape)

model/RIFE2F.py

@@ -34,29 +34,15 @@ def conv_woact(in_planes, out_planes, kernel_size=3, stride=1, padding=1, dilati
                   padding=padding, dilation=dilation, bias=True),
     )
 
-class ResBlock(nn.Module):
+class Conv2(nn.Module):
     def __init__(self, in_planes, out_planes, stride=2):
-        super(ResBlock, self).__init__()
-        if in_planes == out_planes and stride == 1:
-            self.conv0 = nn.Identity()
-        else:
-            self.conv0 = nn.Conv2d(in_planes, out_planes,
-                                   3, stride, 1, bias=False)
+        super(Conv2, self).__init__()
         self.conv1 = conv(in_planes, out_planes, 3, stride, 1)
-        self.conv2 = conv_woact(out_planes, out_planes, 3, 1, 1)
-        self.relu1 = nn.PReLU(1)
-        self.relu2 = nn.PReLU(out_planes)
-        self.fc1 = nn.Conv2d(out_planes, 16, kernel_size=1, bias=False)
-        self.fc2 = nn.Conv2d(16, out_planes, kernel_size=1, bias=False)
-
+        self.conv2 = conv(out_planes, out_planes, 3, 1, 1)
+        
     def forward(self, x):
-        y = self.conv0(x)
         x = self.conv1(x)
         x = self.conv2(x)
-        w = x.mean(3, True).mean(2, True)
-        w = self.relu1(self.fc1(w))
-        w = torch.sigmoid(self.fc2(w))
-        x = self.relu2(x * w + y)
         return x
 
 c = 16
@@ -64,36 +50,32 @@ c = 16
 class ContextNet(nn.Module):
     def __init__(self):
         super(ContextNet, self).__init__()
-        self.conv1 = ResBlock(3, c, 1)
-        self.conv2 = ResBlock(c, 2*c)
-        self.conv3 = ResBlock(2*c, 4*c)
-        self.conv4 = ResBlock(4*c, 8*c)
+        self.conv1 = Conv2(3, c, 1)
+        self.conv2 = Conv2(c, 2*c)
+        self.conv3 = Conv2(2*c, 4*c)
+        self.conv4 = Conv2(4*c, 8*c)
 
     def forward(self, x, flow):
         x = self.conv1(x)
         f1 = warp(x, flow)
         x = self.conv2(x)
-        flow = F.interpolate(flow, scale_factor=0.5, mode="bilinear",
-                             align_corners=False) * 0.5
+        flow = F.interpolate(flow, scale_factor=0.5, mode="bilinear", align_corners=False, recompute_scale_factor=False) * 0.5
         f2 = warp(x, flow)
         x = self.conv3(x)
-        flow = F.interpolate(flow, scale_factor=0.5, mode="bilinear",
-                             align_corners=False) * 0.5
+        flow = F.interpolate(flow, scale_factor=0.5, mode="bilinear", align_corners=False, recompute_scale_factor=False) * 0.5
         f3 = warp(x, flow)
         x = self.conv4(x)
-        flow = F.interpolate(flow, scale_factor=0.5, mode="bilinear",
-                             align_corners=False) * 0.5
+        flow = F.interpolate(flow, scale_factor=0.5, mode="bilinear", align_corners=False, recompute_scale_factor=False) * 0.5
         f4 = warp(x, flow)
         return [f1, f2, f3, f4]
 
-
 class FusionNet(nn.Module):
     def __init__(self):
         super(FusionNet, self).__init__()
-        self.down0 = ResBlock(8, 2*c, 1)
-        self.down1 = ResBlock(4*c, 4*c)
-        self.down2 = ResBlock(8*c, 8*c)
-        self.down3 = ResBlock(16*c, 16*c)
+        self.down0 = Conv2(12, 2*c, 1)
+        self.down1 = Conv2(4*c, 4*c)
+        self.down2 = Conv2(8*c, 8*c)
+        self.down3 = Conv2(16*c, 16*c)
         self.up0 = deconv(32*c, 8*c)
         self.up1 = deconv(16*c, 4*c)
         self.up2 = deconv(8*c, 2*c)
@@ -101,14 +83,14 @@ class FusionNet(nn.Module):
         self.conv = nn.Conv2d(c, 4, 3, 2, 1)
 
     def forward(self, img0, img1, flow, c0, c1, flow_gt):
-        warped_img0 = warp(img0, flow)
-        warped_img1 = warp(img1, -flow)
+        warped_img0 = warp(img0, flow[:, :2])
+        warped_img1 = warp(img1, flow[:, 2:4])
         if flow_gt == None:
             warped_img0_gt, warped_img1_gt = None, None
         else:
             warped_img0_gt = warp(img0, flow_gt[:, :2])
             warped_img1_gt = warp(img1, flow_gt[:, 2:4])
-        s0 = self.down0(torch.cat((warped_img0, warped_img1, flow), 1))
+        s0 = self.down0(torch.cat((img0, img1, warped_img0, warped_img1), 1))
         s1 = self.down1(torch.cat((s0, c0[0], c1[0]), 1))
         s2 = self.down2(torch.cat((s1, c0[1], c1[1]), 1))
         s3 = self.down3(torch.cat((s2, c0[2], c1[2]), 1))
@@ -119,7 +101,6 @@ class FusionNet(nn.Module):
         x = self.conv(x)
         return x, warped_img0, warped_img1, warped_img0_gt, warped_img1_gt
 
-
 class Model:
     def __init__(self, local_rank=-1):
         self.flownet = IFNet()
@@ -158,14 +139,17 @@ class Model:
         self.contextnet.to(device)
         self.fusionnet.to(device)
 
-    def load_model(self, path, rank=0):
+    def load_model(self, path, rank=-1):
         def convert(param):
-            return {
-                k.replace("module.", ""): v
-                for k, v in param.items()
-                if "module." in k
-            }
-        if rank == 0:
+            if rank == -1:
+                return {
+                    k.replace("module.", ""): v
+                    for k, v in param.items()
+                    if "module." in k
+                }
+            else:
+                return param
+        if rank <= 0:
             self.flownet.load_state_dict(
                 convert(torch.load('{}/flownet.pkl'.format(path), map_location=device)))
             self.contextnet.load_state_dict(
@@ -173,21 +157,19 @@ class Model:
             self.fusionnet.load_state_dict(
                 convert(torch.load('{}/unet.pkl'.format(path), map_location=device)))
 
-    def save_model(self, path, rank=0):
+    def save_model(self, path, rank):
         if rank == 0:
-            torch.save(self.flownet.state_dict(),
-                       '{}/flownet.pkl'.format(path))
-            torch.save(self.contextnet.state_dict(),
-                       '{}/contextnet.pkl'.format(path))
+            torch.save(self.flownet.state_dict(), '{}/flownet.pkl'.format(path))
+            torch.save(self.contextnet.state_dict(), '{}/contextnet.pkl'.format(path))
             torch.save(self.fusionnet.state_dict(), '{}/unet.pkl'.format(path))
 
     def predict(self, imgs, flow, training=True, flow_gt=None):
         img0 = imgs[:, :3]
         img1 = imgs[:, 3:]
         flow = F.interpolate(flow, scale_factor=2.0, mode="bilinear",
-	                     align_corners=False) * 2.0
-        c0 = self.contextnet(img0, flow)
-        c1 = self.contextnet(img1, -flow)
+                             align_corners=False) * 2.0
+        c0 = self.contextnet(img0, flow[:, :2])
+        c1 = self.contextnet(img1, flow[:, 2:4])
         refine_output, warped_img0, warped_img1, warped_img0_gt, warped_img1_gt = self.fusionnet(
             img0, img1, flow, c0, c1, flow_gt)
         res = torch.sigmoid(refine_output[:, :3]) * 2 - 1
@@ -201,9 +183,8 @@ class Model:
             return pred
 
     def inference(self, img0, img1):
-        with torch.no_grad():
-            imgs = torch.cat((img0, img1), 1)
-            flow, _ = self.flownet(imgs)
+        imgs = torch.cat((img0, img1), 1)
+        flow, _ = self.flownet(imgs)
         return self.predict(imgs, flow, training=False)
 
     def update(self, imgs, gt, learning_rate=0, mul=1, training=True, flow_gt=None):
@@ -222,10 +203,14 @@ class Model:
                 loss_flow = torch.abs(warped_img0_gt - gt).mean()
                 loss_mask = torch.abs(
                     merged_img - gt).sum(1, True).float().detach()
+                loss_mask = F.interpolate(loss_mask, scale_factor=0.5, mode="bilinear",
+                                          align_corners=False).detach()
+                flow_gt = (F.interpolate(flow_gt, scale_factor=0.5, mode="bilinear",
+                                         align_corners=False) * 0.5).detach()
             loss_cons = 0
             for i in range(3):
-                loss_cons += self.epe(flow_list[i], flow_gt[:, :2], 1)
-                loss_cons += self.epe(-flow_list[i], flow_gt[:, 2:4], 1)
+                loss_cons += self.epe(flow_list[i][:, :2], flow_gt[:, :2], 1)
+                loss_cons += self.epe(flow_list[i][:, 2:4], flow_gt[:, 2:4], 1)
             loss_cons = loss_cons.mean() * 0.01
         else:
             loss_cons = torch.tensor([0])

model/RIFE2F15C.py

@@ -34,29 +34,15 @@ def conv_woact(in_planes, out_planes, kernel_size=3, stride=1, padding=1, dilati
                   padding=padding, dilation=dilation, bias=True),
     )
 
-class ResBlock(nn.Module):
+class Conv2(nn.Module):
     def __init__(self, in_planes, out_planes, stride=2):
-        super(ResBlock, self).__init__()
-        if in_planes == out_planes and stride == 1:
-            self.conv0 = nn.Identity()
-        else:
-            self.conv0 = nn.Conv2d(in_planes, out_planes,
-                                   3, stride, 1, bias=False)
+        super(Conv2, self).__init__()
         self.conv1 = conv(in_planes, out_planes, 3, stride, 1)
-        self.conv2 = conv_woact(out_planes, out_planes, 3, 1, 1)
-        self.relu1 = nn.PReLU(1)
-        self.relu2 = nn.PReLU(out_planes)
-        self.fc1 = nn.Conv2d(out_planes, 16, kernel_size=1, bias=False)
-        self.fc2 = nn.Conv2d(16, out_planes, kernel_size=1, bias=False)
-
+        self.conv2 = conv(out_planes, out_planes, 3, 1, 1)
+        
     def forward(self, x):
-        y = self.conv0(x)
         x = self.conv1(x)
         x = self.conv2(x)
-        w = x.mean(3, True).mean(2, True)
-        w = self.relu1(self.fc1(w))
-        w = torch.sigmoid(self.fc2(w))
-        x = self.relu2(x * w + y)
         return x
 
 c = 24
@@ -64,36 +50,32 @@ c = 24
 class ContextNet(nn.Module):
     def __init__(self):
         super(ContextNet, self).__init__()
-        self.conv1 = ResBlock(3, c, 1)
-        self.conv2 = ResBlock(c, 2*c)
-        self.conv3 = ResBlock(2*c, 4*c)
-        self.conv4 = ResBlock(4*c, 8*c)
+        self.conv1 = Conv2(3, c, 1)
+        self.conv2 = Conv2(c, 2*c)
+        self.conv3 = Conv2(2*c, 4*c)
+        self.conv4 = Conv2(4*c, 8*c)
 
     def forward(self, x, flow):
         x = self.conv1(x)
         f1 = warp(x, flow)
         x = self.conv2(x)
-        flow = F.interpolate(flow, scale_factor=0.5, mode="bilinear",
-                             align_corners=False) * 0.5
+        flow = F.interpolate(flow, scale_factor=0.5, mode="bilinear", align_corners=False, recompute_scale_factor=False) * 0.5
         f2 = warp(x, flow)
         x = self.conv3(x)
-        flow = F.interpolate(flow, scale_factor=0.5, mode="bilinear",
-                             align_corners=False) * 0.5
+        flow = F.interpolate(flow, scale_factor=0.5, mode="bilinear", align_corners=False, recompute_scale_factor=False) * 0.5
         f3 = warp(x, flow)
         x = self.conv4(x)
-        flow = F.interpolate(flow, scale_factor=0.5, mode="bilinear",
-                             align_corners=False) * 0.5
+        flow = F.interpolate(flow, scale_factor=0.5, mode="bilinear", align_corners=False, recompute_scale_factor=False) * 0.5
         f4 = warp(x, flow)
         return [f1, f2, f3, f4]
 
-
 class FusionNet(nn.Module):
     def __init__(self):
         super(FusionNet, self).__init__()
-        self.down0 = ResBlock(8, 2*c, 1)
-        self.down1 = ResBlock(4*c, 4*c)
-        self.down2 = ResBlock(8*c, 8*c)
-        self.down3 = ResBlock(16*c, 16*c)
+        self.down0 = Conv2(12, 2*c, 1)
+        self.down1 = Conv2(4*c, 4*c)
+        self.down2 = Conv2(8*c, 8*c)
+        self.down3 = Conv2(16*c, 16*c)
         self.up0 = deconv(32*c, 8*c)
         self.up1 = deconv(16*c, 4*c)
         self.up2 = deconv(8*c, 2*c)
@@ -101,14 +83,14 @@ class FusionNet(nn.Module):
         self.conv = nn.Conv2d(c, 4, 3, 2, 1)
 
     def forward(self, img0, img1, flow, c0, c1, flow_gt):
-        warped_img0 = warp(img0, flow)
-        warped_img1 = warp(img1, -flow)
+        warped_img0 = warp(img0, flow[:, :2])
+        warped_img1 = warp(img1, flow[:, 2:4])
         if flow_gt == None:
             warped_img0_gt, warped_img1_gt = None, None
         else:
             warped_img0_gt = warp(img0, flow_gt[:, :2])
             warped_img1_gt = warp(img1, flow_gt[:, 2:4])
-        s0 = self.down0(torch.cat((warped_img0, warped_img1, flow), 1))
+        s0 = self.down0(torch.cat((img0, img1, warped_img0, warped_img1), 1))
         s1 = self.down1(torch.cat((s0, c0[0], c1[0]), 1))
         s2 = self.down2(torch.cat((s1, c0[1], c1[1]), 1))
         s3 = self.down3(torch.cat((s2, c0[2], c1[2]), 1))
@@ -119,7 +101,6 @@ class FusionNet(nn.Module):
         x = self.conv(x)
         return x, warped_img0, warped_img1, warped_img0_gt, warped_img1_gt
 
-
 class Model:
     def __init__(self, local_rank=-1):
         self.flownet = IFNet()
@@ -158,14 +139,17 @@ class Model:
         self.contextnet.to(device)
         self.fusionnet.to(device)
 
-    def load_model(self, path, rank=0):
+    def load_model(self, path, rank=-1):
         def convert(param):
-            return {
-                k.replace("module.", ""): v
-                for k, v in param.items()
-                if "module." in k
-            }
-        if rank == 0:
+            if rank == -1:
+                return {
+                    k.replace("module.", ""): v
+                    for k, v in param.items()
+                    if "module." in k
+                }
+            else:
+                return param
+        if rank <= 0:
             self.flownet.load_state_dict(
                 convert(torch.load('{}/flownet.pkl'.format(path), map_location=device)))
             self.contextnet.load_state_dict(
@@ -173,21 +157,19 @@ class Model:
             self.fusionnet.load_state_dict(
                 convert(torch.load('{}/unet.pkl'.format(path), map_location=device)))
 
-    def save_model(self, path, rank=0):
+    def save_model(self, path, rank):
         if rank == 0:
-            torch.save(self.flownet.state_dict(),
-                       '{}/flownet.pkl'.format(path))
-            torch.save(self.contextnet.state_dict(),
-                       '{}/contextnet.pkl'.format(path))
+            torch.save(self.flownet.state_dict(), '{}/flownet.pkl'.format(path))
+            torch.save(self.contextnet.state_dict(), '{}/contextnet.pkl'.format(path))
             torch.save(self.fusionnet.state_dict(), '{}/unet.pkl'.format(path))
 
     def predict(self, imgs, flow, training=True, flow_gt=None):
         img0 = imgs[:, :3]
         img1 = imgs[:, 3:]
         flow = F.interpolate(flow, scale_factor=2.0, mode="bilinear",
-	                     align_corners=False) * 2.0
-        c0 = self.contextnet(img0, flow)
-        c1 = self.contextnet(img1, -flow)
+                             align_corners=False) * 2.0
+        c0 = self.contextnet(img0, flow[:, :2])
+        c1 = self.contextnet(img1, flow[:, 2:4])
         refine_output, warped_img0, warped_img1, warped_img0_gt, warped_img1_gt = self.fusionnet(
             img0, img1, flow, c0, c1, flow_gt)
         res = torch.sigmoid(refine_output[:, :3]) * 2 - 1
@@ -201,9 +183,8 @@ class Model:
             return pred
 
     def inference(self, img0, img1):
-        with torch.no_grad():
-            imgs = torch.cat((img0, img1), 1)
-            flow, _ = self.flownet(imgs)
+        imgs = torch.cat((img0, img1), 1)
+        flow, _ = self.flownet(imgs)
         return self.predict(imgs, flow, training=False)
 
     def update(self, imgs, gt, learning_rate=0, mul=1, training=True, flow_gt=None):
@@ -222,10 +203,14 @@ class Model:
                 loss_flow = torch.abs(warped_img0_gt - gt).mean()
                 loss_mask = torch.abs(
                     merged_img - gt).sum(1, True).float().detach()
+                loss_mask = F.interpolate(loss_mask, scale_factor=0.5, mode="bilinear",
+                                          align_corners=False).detach()
+                flow_gt = (F.interpolate(flow_gt, scale_factor=0.5, mode="bilinear",
+                                         align_corners=False) * 0.5).detach()
             loss_cons = 0
             for i in range(3):
-                loss_cons += self.epe(flow_list[i], flow_gt[:, :2], 1)
-                loss_cons += self.epe(-flow_list[i], flow_gt[:, 2:4], 1)
+                loss_cons += self.epe(flow_list[i][:, :2], flow_gt[:, :2], 1)
+                loss_cons += self.epe(flow_list[i][:, 2:4], flow_gt[:, 2:4], 1)
             loss_cons = loss_cons.mean() * 0.01
         else:
             loss_cons = torch.tensor([0])

model/warplayer.py

@@ -19,4 +19,4 @@ def warp(tenInput, tenFlow):
                          tenFlow[:, 1:2, :, :] / ((tenInput.shape[2] - 1.0) / 2.0)], 1)
 
     g = (backwarp_tenGrid[k] + tenFlow).permute(0, 2, 3, 1)
-    return torch.nn.functional.grid_sample(input=tenInput, grid=torch.clamp(g, -1, 1), mode='bilinear', padding_mode='zeros', align_corners=True)
+    return torch.nn.functional.grid_sample(input=tenInput, grid=g, mode='bilinear', padding_mode='border', align_corners=True)