Face Attribute Prediction on CelebA benchmark with PyTorch Implemantation, heavily borrowed from my MobileNetV2 implementation.
- Anaconda3 (Python 3.6+, with Numpy etc.)
- PyTorch 0.4+
- tensorboard, tensorboardX
CelebA dataset is a large-scale face dataset with attribute-based annotations. Cropped and aligned face regions are utilized as the training source. For the pre-processed data and specific split, please feel free to contact me: [email protected]
- Both ResNet and MobileNet as the backbone for scalability
- Each of the 40 annotated attributes predicted with multi-head networks
- Achieve ~92% average accuracy, comparative to state-of-the-art
- Fast convergence (5~10 epochs) through finetuning the ImageNet pre-trained models