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
361 stars 76 forks source link

training numbers in test set #54

Open felixchengxh opened 2 years ago

felixchengxh commented 2 years ago

Hi, thanks for your great work. I have a little question, why does the number of training images in test_split is 7397? Seems like you moved 1% of training images to validation set instead of using the whole training set, so I wonder what's the difference and why you do so? Thank you.

Owen-Liuyuxuan commented 2 years ago

It is like a mini-validation set before submission. You can train with all the images but it receives similar results.

felixchengxh commented 2 years ago

Another question please, I had a little modify on your model, when I submit my data on KITTI benchmark, only 2d detection results can be received, do you know what made this happen? image

Owen-Liuyuxuan commented 2 years ago

Have you waited enough time for the results to pop up? Because if you strictly follow my code, bird-eye-view and 3D results will come up together. Only BEV presented is weird.

felixchengxh commented 2 years ago

Yes, I did, the result still remain 0 until now(I tested like 10 hours ago), and the 3D results didn't show, is there anything I can modify in the code to avoid this?