We release various convolutional neural networks (CNNs) trained on Human Rights Archive (HRA) to the public. HRA is the first verified-by-experts repository of human rights violations photographs, labelled with human rights semantic categories, comprising a list of the types of human rights abuses encountered at present.
- Baseline model
- ResNet50-HRA
- VGG16-HRA
- VGG16-Places365-HRA
- VGG19
- CompoundNet-ResNet50
- CompoundNet-VGG16
- CompoundNet-VGG19
Train Human-Rights-Archive-CNNs using Keras. The training script is here.
Example
--------
>>> python train_HRA.py --pre_trained_model VGG16 --pooling_mode avg
The performance of the CNNs is listed below.
Some qualitative prediction results using the VGG16-Places365:
