857 questions
Best practices
0
votes
0
replies
39
views
FID score computation while training WGAN-GP
I am training WGAN-GP on Eurosat dataset, splitted into train/val/test sets in counts 18900/4050/4050. Since FID scores are widely used in GANs in image generation, I based my hyperparameter search on ...
0
votes
0
answers
45
views
Trained and loaded CycleGAN model is giving distorted output images
I trained a CycleGAN model on Google Colab using this repository - https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix
The model should enhance dark images. I tested the model on my test dataset ...
- 11
Tooling
0
votes
6
replies
134
views
Synthetic to real chess images project
I have a project in university to generate a real chess image from a rendered image.
I have 520 pairs of images but there is no exact match pixel to pixel; it's close to that, plus there an third ...
- 1
0
votes
1
answer
122
views
ImportError: cannot import name 'Generator' from 'model' (unknown location) in GAN training script
I'm running a GAN training pipeline in Google Colab. The preprocessing scripts run without any issues:
GitHub repository
!python run_preprocess.py --dataset custom --train_num 500
!python ...
- 1
0
votes
2
answers
133
views
Understanding PyTorch gradients and backward function, when backwarding more than once
I'm trying to add more than one generator training step per cycle to a GAN, i.e. I want my Generator to update its parameters n times every m updates of the Discriminator, where n > m.
I wrote this ...
- 5
0
votes
0
answers
65
views
PyTorch Gan training: shape mismatch between real and fake tensor
I’m working on a custom CTGAN-like model for tabular data, where I implemented my own _collate_fn and run_step() logic to control sampling, noise injection, and discriminator inputs.
However, I ...
0
votes
0
answers
71
views
GAN results fluctuating
I have the following GAN architecture:
import torch, os, torchvision
import torch.nn as nn
import torch.optim as optim
from torchvision import transforms, datasets, utils
class MappingNetwork(nn....
- 3,213
0
votes
0
answers
130
views
Can I add my own function in nn.Sequential?
I'm a beginner in machine learning. Recently, I'm trying to use CGAN for image generation and lebelling.
Currently, I'm manually doing circular padding in convolution operation as following.
def ...
- 21
1
vote
0
answers
61
views
Convolutional Generative Adversarial Networks Noise Patterns
I am coding a DCGAN to produce Brain MRI data, based on the BRATs 2020 dataset. As a sanity check, I am training on a SINGLE image with CONSTANT noise, to see if there are any inherent flaws in my ...
- 313
0
votes
0
answers
47
views
Problems with setting gradients of GAN in keras
Having some experience in developing neural networks in Keras, I decided to write a non-standard GAN, which you can't really call as such. The point is that the discriminator is a ready-made neural ...
- 1
0
votes
2
answers
684
views
Issue with python/keras implementation of Generative Adversarial Network (GAN) on MacBook
I'm very new to keras and Generative Adversarial Networks (GAN). When I attempted to run this script for a simple GAN on my MacBook, I got:
Traceback (most recent call last):
File "~/Library/...
- 27
0
votes
1
answer
92
views
Creating different synthetic Body Shapes for one specific Human Pose for Training Pose classification
I'm working on my master's thesis, which involves tracking singing poses to provide feedback on posture. Specifically, I am using pose images, such as those derived from keypoint detection models like ...
- 21
1
vote
0
answers
57
views
Performance Optimization in DeblurGANv2
---
project: deblur_gan
experiment_desc: 11_50_fpn_stop
train:
# files_a: &FILES_A /datasets/my_dataset/**/*.jpg
#可以使用绝对路径
#files_a: &FILES_A D:/deblur/goprol_large/**/*.png
#files_a: &...
- 11
0
votes
1
answer
47
views
DCGAN loss not making sense to me
I trained a DCGAN model on a dataset. Few things have been observed:
Repetitive patterns in discriminator loss (d_loss = d_loss_real + d_loss_fake) and g_loss
d_loss_real (discriminator loss on real ...
- 420
0
votes
1
answer
95
views
How can the dimensions of a Conv2D layer be calculated?
I'm trying to understand the dimensions of the output of the Generator of my GAN. The dims of the result after each layer is as follows:
Start: torch.Size([128, 74, 1, 1])
After block1: torch.Size([...
- 1