Questions tagged [generative-adversarial-network]
Generative adversarial networks (GANs) are a class of artificial intelligence algorithms used in unsupervised (and semi-supervised) machine learning, implemented by a system of two neural networks contesting with each other in a zero-sum game framework.
generative-adversarial-network
918
questions
0
votes
0
answers
14
views
CGAN Training Issues: Discriminator Accuracy at 100% and Generator Loss at 0
I am trying to train a Conditional Generative Adversarial Network (CGAN) to generate synthetic leaf images. However, during training, my discriminator's accuracy quickly reaches 100%, and the ...
-1
votes
0
answers
22
views
Fraud detection using GAN
I am implementing a fraud detection model based on transactions using GAN but I still want to specify my model, i.e. I want to emphasize the RIB and transaction time (and especially issue time) I want ...
0
votes
0
answers
9
views
OperatorNotAllowedInGraphError when trying to set discriminator.trainable property
I'm training a GAN based on the pix2pix architecture, and my training code works in Eager mode. Putting it in graph mode returns the following error:
OperatorNotAllowedInGraphError: Using a ...
0
votes
0
answers
10
views
How to restart a GAN Training with TensorFlow 2.15 using checkpoints
I'm working on creating a Jupyter notebook for training a GAN using TensorFlow, and I want to be able to restore the last checkpoint to continue training from where I left off.
I am following this ...
0
votes
0
answers
14
views
Inconsistent Inception Scores for DCGAN with Same Data and Noise - What Could Be the Issue?
I've been working on training a DCGAN on the CIFAR-10 dataset, and I'm evaluating the quality of generated images using the Inception Score. However, I've noticed that the Inception Scores vary ...
0
votes
0
answers
28
views
Stability problem in training GAN (D loss and G loss)
I've just get acquainted with GAN model. I decided to build a DCGAN model base on this link, the different is that they use keras and i use pytorch.
The problem is my D loss and my G loss seems not ...
0
votes
0
answers
10
views
RuntimeError: mat1 and mat2 shapes cannot be multiplied (128x256 and 32768x1)
class Discriminator(nn.Module):
def __init__(self, img_shape):
super(Discriminator, self).__init__()
self.Conv1 = nn.Conv2d(3, 16, 4, stride=2, padding=1)
self....
0
votes
0
answers
27
views
Confusion about output sizes of GAN
I am trying to understand a code, I am confused about the test cells. When i am printing the shape of the output it is hidden_output.shape =(num_test, 20, 4, 4), test_hidden_block_stride(hidden_output)...
0
votes
0
answers
20
views
Latent vectors that correspond to race and gender in StyleGAN3
As a part of my research, I need to produce 4 sets of fictitious profile image that are identical in all features of the image, except race and gender (White male, White female, Black male, Black ...
0
votes
0
answers
9
views
How to do dynamic upscaling using ESPCN?
I am currently doing 2x and 3x upscaling using ESPCN using pre-defined weights, is there any way that I can do dynamic upscaling, by changing the parameters/models/weights, basically anything?
I am ...
1
vote
0
answers
30
views
Image inference with cycleGan model from CycleGAN-and-pix2pix open source repo
I have trained a cycleGan model on google colab according to the CycleGAN-and-pix2pix opensource API.
For the train process I used !python train.py --dataroot /content/drive/MyDrive/project/dataset --...
0
votes
1
answer
34
views
Blending bitmap tiles
I'm using a GAN tflite model to process a bitmap. T model only accepts images of 512 x 512 pixels, so I first crop the image into tiles and then process all tiles at once. This resulted in an image ...
-2
votes
0
answers
26
views
import error: 'bytes' object cannot be interpreted as an integer
I am trying to generate synthetic data using Chronos in the following link.
https://analytics-zoo.readthedocs.io/en/latest/doc/Chronos/Overview/chronos.html#generate-synthetic-data
However, the import ...
0
votes
0
answers
29
views
Imbalance Between Generator and Discriminator Losses in GAN Training for Super-Resolution
I am training a Generative Adversarial Network (GAN) for the task of image super-resolution. However, I am encountering a persistent issue where the generator's loss decreases significantly over time, ...
0
votes
0
answers
28
views
Execution is very slow for GAN
I wrote a code on GAN and had previously trained in the Colab environment. I leave the Colab code below.
def build_generator(latent_dim):
"""Build the generator model.""&...