We have reproduced mAP on TAO dataset, and found that the performance is quite low. | AP | AP50 | AP75 | track3D mAP | | :--: | :--: | :--: | :--: | | 8.8717 | 14.0362 | 9.4615| 7.7657 | The numbers are generated with the block [here](https://github.com/SysCV/tet/blob/main/teter/datasets/tao_dataset.py#L240) , while track3D mAP is calculated with the [tao toolkit](https://github.com/TAO-Dataset/tao). Could you provide the mAP numbers you get on TAO? We need to further verify it there is any bugs here