Although our method is designed for open-world recognition, we also provide the code for experiments on ImageNet for reference.
Change the ImageNet dataset path in tools/create_loading_list.py
and run:
cd imagenet/tools && python3 create_loading_list.py
./scripts/run train.py configs/config_in1k.py
./scripts/run eval_hook.py evals/eval.py configs/config_in1k.py
params | training group | R | P | F1 |
---|---|---|---|---|
1.78B | ImageNet-1K | 0.896 | 0.432 | 0.574 |
The results are obtained by training the model on ImageNet-1K only and evaluating on the validation set.