Danila Rukhovich

Can you please attach the full log of this run?

Does this [comment](https://github.com/SamsungLabs/tr3d/blob/main/mmdet3d/core/bbox/structures/depth_box3d.py#L11) help you? For the yaw angle it says that `yaw=0` when it is oriented as `x` axis. And the center of the box is not in its...

The prediction of angle may be not quite accurate. And also we don't make much difference between rotation on 90 degrees. You can see more details on angle question in...

Can you print the shapes of `calib['K']` and `rt_mat`? I think they both should be `(3, 3)`, as we save them [here](https://github.com/SamsungLabs/tr3d/blob/main/tools/data_converter/sunrgbd_data_utils.py#L121).

> Why do I fail to scan the scan dataset when running Python batchreload_scannet_data.py can you please share the full error log?

may you try to reshape 9 to (3, 3)? btw did you run this [script](https://github.com/SamsungLabs/tr3d/blob/main/tools/data_converter/sunrgbd_data_utils.py#L121)? are both matrices (3, 3) after it?

Looks like the real error is hidden by this [exception](https://github.com/SamsungLabs/tr3d/blob/main/data/scannet/batch_load_scannet_data.py#L106). Can you please share the error trace at this line?

Have you downloaded the scannet dataset and put it to this `scans/` directory? The first item in the instruction [here](https://github.com/SamsungLabs/tr3d/tree/main/data/scannet).

We save both gt and predictied boxes for visualization with this [function](https://github.com/SamsungLabs/tr3d/blob/main/mmdet3d/core/visualizer/show_result.py#L77). Resulting `.obj` files can be opened with the meshlab together with point cloud `.obj` file.

To reduce the number of predicted boxes please follow our advice about `score_thr` in the readme. Gt boxes looks pretty much fine. May be you can try it on scene...