search

zhaoweicai / mscnn

Caffe implementation of our multi-scale object detection framework
404 stars 211 forks source link

The effect is lower than expected #62

Closed qianyeqiang closed 7 years ago

qianyeqiang commented 7 years ago

Dear author I test the KITTI vehicle training dataset using the mscnn-8s-768-trainval-pretrained, which is download as README. Then result is evaluated. The result is 77% (moderate), which is far below the result in KITTI list. Since the training set and the test set are the same, the result should be close to 100% from my experience. So, what the problem? Thank you very much. @zhaoweicai

qianyeqiang commented 7 years ago

I have find the problem, I change the detection code incorrectly. I will close the issue.