Projects
Built this: Modern DCGAN reproducibility study
A DCGAN (Radford et al., 2015) reproduction that keeps the all-convolutional architecture but swaps in LSGAN loss, spectral normalisation and label smoothing, plus an analysis toolkit for filters and feature maps.
- Same architecture as the 2015 paper, more stable training engine: LSGAN loss, spectral norm, label smoothing and discriminator dropout.
- A GANAnalyzer class hooks into layers to visualise learned filters and feature maps without retraining.