Danila Rukhovich
Danila Rukhovich
## Motivation Fix #1922. Following [lilanxiao/Rotated_IoU#39#issuecomment-1146352088](https://github.com/lilanxiao/Rotated_IoU/issues/39#issuecomment-1146352088) we remove `cuda` and `cpp` part of `diff_iou_rotated` in favour of pure `pytorch` implementation. It has negligible affect on speed however the overall accuracy...
Looks like [here](https://github.com/apple/ARKitScenes/blob/main/threedod/benchmark_scripts/utils/tenFpsDataLoader.py#L136) we need to replace `len(gt) == 0` with `len(gt) == 0 or len(gt.get('data', [])) == 0` otherwise for scene `42897846` we get `ValueError: need at least one...
``` import cadquery as cq cq.Workplane().sketch().push([(48.5800789794081,-3.3737790291878085)]).rect(3.0,87.6847679284009).rect(3.0,88.65465934289723).polygon([(0,0),(1,1),(2,2)],mode="s").finalize().extrude(1) ``` cadquery==2.4.0 is installed throw pip
Hi, I think [this](https://github.com/XavierJiezou/DiffCR/blob/main/data/dataset.py#L135) line is basically not ok.
More discussion in [#696#issuecomment-874275857](https://github.com/open-mmlab/mmdetection3d/pull/696#issuecomment-874275857). It's WIP now as we also need a unit test and probably ScanNet + `MultiViewPipeline` test. Can I add 2-3 ScanNet .jpg images to `tests/data` for...