Download the dataset from here: dataset
Download the pretrained VGG weights here: pretrained VGG weights
Download the Faster RCNN model weights with FC-net here for testing purposes: FC-net + RPN weights
Download the Faster RCNN model weights with VGG-net here for testing purposes: VGG-net + RPN weights
python train_frcnn.py --path="dataset/png_TrainIJCNN2013/gt_train.txt" --network='vgg' --hf=True --vf=True --rot=True --num_epochs=10 --epoch_length=600 --config_filename='vggnet_config.pickle' --output_weight_path='model_frcnn_vgg.hdf5'
python train_frcnn.py --path="dataset/png_TrainIJCNN2013/gt_train.txt" --network=‘fcnet’ --hf=True --vf=True --rot=True --num_epochs=10 --epoch_length=600 --config_filename=‘fcnet_config.pickle' --output_weight_path='model_frcnn_fcnet.hdf5'
The results (images with bounding boxes) will be saved in results_imgs folder
python test_frcnn.py --path='dataset/png_TestIJCNN2013/gt_test.txt' --config_filename='vggnet_config.pickle' --network='vgg'
python test_frcnn.py --path='dataset/png_TestIJCNN2013/gt_test.txt' --config_filename='fcnet_config.pickle' --network='fcnet'