bennyustc/monocular-camera-3d-object-detection

3D object detection using monocular camera

Monocular-camera-3D-object-detection

The yolo model is taken from the following : https://github.com/zzh8829/yolov3-tf2

To download and Convert pre-trained Darknet weights in the above repo

yolov3

wget https://pjreddie.com/media/files/yolov3.weights -O data/yolov3.weights
python convert.py --weights ./data/yolov3.weights --output ./checkpoints/yolov3.tf

Samples of the Detection algorithim

The algorithim

1. Yolov3 algorithim is used to draw 2D bounding boxes around the object
   
2. Another deep learning model is used to get the orientation of each detected object
   
3. Following the geometric constraints equations from this paper: https://arxiv.org/abs/1612.00496, one could calculate the displacement between the camera and the center of the object.
   
4. We assume that the dimention of any object is the average dimention of the object's class in the kitti datset

references:

https://arxiv.org/abs/1612.00496
https://github.com/zzh8829/yolov3-tf2