Below is the architectural diagram of DCGAN
Trained DCGAN on STL10 dataset. Each image is of size 3x96x96. Trained DCGAN for 50 Epochs on T4x2 GPU's and batch size of 128.
Below are the results. While results may not be very clear due to limited training, still we can see some sembelence of objects and figures.
Trained DCGAN on CIFAR-10 dataset. Each image is of size 3x32x32. Trained DCGAN for 100 Epochs on T4x2 GPU's and batch size of 256.
Below are some results straight out of Generator
