about eval and test

[xuchen-dev]

hi i find validation during training result using kitti_kitti_validation is better than test using kitti-zhou-test

i wonder is kitti-zhou-test result is your paper provide or kitti-kitti-validation or other like kitti-zhou-validation

[xuchen-dev]

kitti-zhou-validation looks even better

[klingner]

I am not quite sure what you are aiming at. Results on different datasets are of course different and not comparable. A better result on one dataset mainly tells you that the dataset is easier for the model. In the paper mainly results on kitti_zhou_test are reported as is common.

[xuchen-dev]

thank you