Tianyu Li

I had the same problem when calling `metrics_module.detection_metrics`. My version of pip package is 1.4.3 with tf 2.6.0. I don't see the interface changed in metrics_ops.cc, it seems that `metrics_ops.so`...

I am experiencing the same issue. It appears that when the viewer renders the image, the model is not correctly set to eval mode.

You can also use the key `is_intersection_or_connector` to filter the lane segments within the intersection.

Hello! The code in this repository builds upon last year's [baseline](https://github.com/CVPR2023-3D-Occupancy-Prediction/CVPR2023-3D-Occupancy-Prediction), which deliberately omitted the temporal component of BEVFormer. It appears that the unresolved issue stems from BEVFormer's shift calculations...

We have a fix about the coordinates issue at LaneSegNet repo. You can check it out [here](https://github.com/OpenDriveLab/LaneSegNet/blob/main/projects/lanesegnet/models/modules/bevformer_constructer.py#L122).

The OpenOcc dataset doesn't construct a can_bus tensor. You can canstruct one or modify the code as you like.

Here is a [doc](https://github.com/fundamentalvision/BEVFormer/blob/master/docs/can_bus.ipynb) for can bus.

Yes, results may degrade when the ego rotates. Consider using BEV rotation augmentation to enhance performance. For the visualization of BEV feature, you can try using the color map to...

Thanks for your question. You may can upgrade your PyTorch version to fix this. The repository's dependency is flexible.

I have a combination with torch 2.0 and CUDA 11.7/11.8, feel free to try. I think torch 1.13 can also compatible with mmcv-full 1.6.0. You can directly try on 1.13....