The evaluation code of BEV is integrated into simple-romp, evaluation folder.
To run the evaluation, please first install simple-romp following the guidance.
To prepare the evaluation datasets, please refer to dataset.md.
In each evaluation code, please set the dataset path and output_save_dir for saving the predictions.
Please properly set the path in simple_romp/evaluation/eval_Relative_Human.py and then run
cd simple_romp/evaluation
python eval_Relative_Human.py
You can also download the predictions.
Please properly set the path in simple_romp/evaluation/eval_AGORA.py and then run
cd simple_romp/evaluation
python eval_AGORA.py
Please properly set the path in simple_romp/evaluation/eval_cmu_panoptic.py and then run
cd simple_romp/evaluation
python eval_cmu_panoptic.py
You can also download the predictions and then set the results path in evaluation_results function to get the matrix numbers.