Closed mcanan closed 5 months ago
I tried it on my Windows 10 computer and it seems to work well. You can see the following information:
[INFO] LogHandlers setup!
22-07-28 14:07:53.993 : ********************* Validation Phase ************************
22-07-28 14:07:53.994 : => Import libs
22-07-28 14:07:54.012 : => Set validation data
22-07-28 14:07:54.015 : => Make models and load checkpoint
22-07-28 14:07:56.518 : Loading pytorch model from `Models/pretrain/ProGAN_pt_mixtureData_GAN_MSGAN_ACGAN.pth`.
22-07-28 14:07:56.652 : Successfully loaded!
22-07-28 14:07:56.653 : `lod` of the loaded model is 0.0.
22-07-28 14:07:58.170 : ===> Use 1 GPUs
22-07-28 14:07:58.276 : ===> Load ckp for <Mapping_net> from E:\\code\\baby_model_pool\\mapping\\Mapping_Xencoder_full_ProGAN_GAN_MSGAN_ACGAN_deepArch_multi-gt_v4\\MappingNet_Batchsize_32_Epoch_298.pth
22-07-28 14:07:58.528 : => Begin to validate the model
But I found that I obtain a similar result with you on a Ubuntu computer. I am not sure the reason currently. I will try to figure it out.
Okay it seems that the ProGAN's weights provided are not correct.
Enter the code
folder. You can simply do the following operations and re-run the main.py:
mkdir ./babymapping_1219/Models/pretrain
mv ./baby_model_pool/ProGAN-ckp/* ./babymapping_1219/Models/pretrain/
Hope this can address your issue.
Great! Is working nice now! Thank you.
I followed the steps in 3.3. Testing a Real Face but I am getting as output 320 gray images like this:
I am using the downloaded weights and the default input images defined in the validation.yaml file.
Thanks