fyp-long-tail-recognition

Update notifications

• 03/04/2021: Configured training settings for Places_LT
• 01/05/2021: Update finetuning with knowledge transfer

Requirements

• PyTorch (tested with version == 1.7.1)

Data Preparation

• Download the ImageNet_2014 and Places_365 (256x256 version). Change the data_root in main.py to your own directory.

• CIFAR dataset will be automatically downloaded when running CIFAR experiment.

Getting Started (Training & Testing)

ImageNet-LT

• Baseline (with k=1)

python main.py --m_from 1

• cRT (Decoupling)

python main.py --config ./config/ImageNet_LT/stage_1_crt.py --path <folder path> --crt

• LLW:

python main.py --trainable_logits_weight --path <baseline checkpoint directory name> --scaling_logits --config ./config/ImageNet_LT/weight_finetune_resnet10.py --merge_logits

• Knowledge Transfer Loss:

python main.py --second_dotproduct --m_from 1 --scaling_logits --second_head_alpha 0.2 --temperature 1