Our work proposes a novel Distribution-aware Forgetting Compensation (DAFC) model that explores cross-domain shared representation learning and domain-specific distribution integration without relying on old exemplars or knowledge distillation.
- Python 3.7
- torch 1.10.0
- For more detailed requirements, run
pip install -r requirements.txt
- Please follow Torchreid_Datasets_Doc to download datasets and unzip them to your data path .
- Prepare the Seen dataset structure as follow: ./docs/seen dataset structure .md
- Prepare the Unseen dataset structure as follow: ./docs/Unseen dataset structure .md
Training order-1 for training.
python lifelong_train.py --data-dir=/your seen dataset path --logs-dir==/save path --orders="order-1"
Training order-2 for training.
python lifelong_train.py --data-dir=/your seen dataset path --logs-dir==/save path --orders="order-2"