aihgf / orderless-rnn-classification

Orderless Recurrent Models for Multi-label Classification

orderless-rnn-classification

This repository provides the MS-COCO training code for the Orderless Recurrent Models for Multi-label Classification paper which will be published in CVPR2020.

Environment

• python 2.7
• pytorch 0.4.1

Training

Three steps of training:

Encoder

python train_bce.py -image_path {image_path} -save_path {save_path1}

Encoder + decoder

python train_lstm.py -image_path {image_path} -save_path {save_path2} -order_free pla -finetune_encoder -swa_params "{'lr_high': 1e-3, 'lr_low': 1e-6, 'cycle_length': 3, 'swa_coeff': 0.1}" -encoder_weights {save_path1}/BEST_checkpoint.pt

Decoder

python train_lstm.py -image_path {image_path} -save_path {save_path3} -order_free pla -decoder_lr 1e-5 -snapshot {save_path2}/BEST_checkpoint.pth.tar -epochs 5 -train_from_scratch

Testing

python train_lstm.py -image_path {image_path} -snapshot {save_path3}/BEST_checkpoint.pth.tar -test_model

Acknowledgements

The encoder-decoder architecture that is implemented in this repository is based on the implementation in here.