train 2D detection part only

HKUST-Aerial-Robotics / Stereo-RCNN

Code for 'Stereo R-CNN based 3D Object Detection for Autonomous Driving' (CVPR 2019)

MIT License

690 stars 177 forks source link

Open moragab1993 opened 4 years ago

moragab1993 commented 4 years ago

hello

we tested our own dataset and we noticed that the 2D detection is not good

so is it possible to train the 2D detection part only ?