search

NVIDIA / vid2vid

Pytorch implementation of our method for high-resolution (e.g. 2048x1024) photorealistic video-to-video translation.
Other
8.57k stars 1.2k forks source link

Using madel on my own datasets,the generated img in the results/ folder is quiet different from the generate image in the checkpoints/web/image/ #145

Open birdflyto opened 4 years ago

birdflyto commented 4 years ago

25203520 epoch001_fake_image fake_B_40205480 real_A_40205480 the first image is the real image the second image is the generate image in the checkpoints/web/image/ the 3rd image is the fake image gennerated by running test.py in the results folder the 4th image is the generate real generated by running test.py in the results folder

50

hello,thanks for sharing,when i train the model and test on my own datasets,i can see the generate image in the checkpoints/web/image/ is similar to the real input image,but when i run test.py with tesh.sh guid,the generate fake image is quiet different from the img in checkpoints/web/image/,i try to read all the code in model.py train.py .etc,but i have not found where the key code that i can change to get the imgs in the result like that in checkpoints/web/image/. Many thanks!! @jiaxianhua @swarren @aaronp24 @lukeyeager @drossetti

jiangzhubo commented 4 years ago

@birdflyto i also have this issue, do you solve it?

navuboy commented 3 years ago

@birdflyto where you able to solve this issue?