zhixinlai/3d-detection-with-monocular-rgb-image

3D detection prediction in Autopilot scene with monocular RGB image. Selected Faster RCNN as basemodel and ResNet as backbone with Python.

3d-detection-with-monocular-rgb-image's Issues

extraction of 3D BBOX from 2D BBOX

Thanks for sharing,
I want to know something i have detected mask of an image and both are calibrated. now i can project the mask on image in form of 2d boxes but I want projection in the form of 3d bounding boxes is it possible to estimate 3D bboxes from 2D ? Below is the mask and detected projection.
looking for your help...

How does cal_theta_ray() function work?

Hi, Thanks for your sharing.
I'm quite confused about the function cal_theta_ray(). I have seen it in another project, https://github.com/skhadem/3D-BoundingBox. I know its input are a cropped image, a 2D bbox and P_rect_02 matrix, and its output is theta_ray in figure 3 of your README. But I can't understand these operations in this function.
Do I miss some basic mathematical knowledges? Looking forward for your reply.