search

Owen-Liuyuxuan / visualDet3D

Official Repo for Ground-aware Monocular 3D Object Detection for Autonomous Driving / YOLOStereo3D: A Step Back to 2D for Efficient Stereo 3D Detection
https://owen-liuyuxuan.github.io/papers_reading_sharing.github.io/3dDetection/GroundAwareConvultion/
Apache License 2.0
362 stars 77 forks source link

Results about validation set #34

Closed oliver0922 closed 2 years ago

oliver0922 commented 2 years ago

Hi LIU

I appreciate for your work.

I am curious about the results of validation set.

image

I got these results while training. Would you please explain the results of the image above?

Thanks

Owen-Liuyuxuan commented 2 years ago

It looks like you are using the "test_split", as the default in stereo3D_example.

The validation set of the "test_split" contains only a few images extracted from the 7400+ training set. The training set of the "test_split" contains almost all the 7400+ training set. Models trained on this split should evaluate on the test set, and uploaded to the kitti server.

If you want to train with "chen_split" and conduct validation completely in the local machines, you need to change the split file to "chen_split", and then re-run all the pre_compute and training scripts.

oliver0922 commented 2 years ago

Thanks for your answer.

What I am curious about is how to interpret the results

Thanks.

Owen-Liuyuxuan commented 2 years ago

1221285300

You can refer to the official benchmark/devkit for a detailed/simpler c++ implementation