comparision between AVOD and MV3D.

AVOD → Aggregated view object detection.

MV3D → Multi View Object Detection for the autonomous driving.

AVOD provides the better performance even on the smaller objects with the substantial increase in the MAP value.

Both MV3D and AVOD are inspired by the faster rcnn. But the AVOD is improved in many ways.

1. MV3D uses only bird eye view for the proposal generation. Where as the AVOD uses the both bird eye view and the RGB images for the proposal generation.
2. AVOD also improves the MV3D by improved box encoding, heading angle regression, using FPN for the detection of the small objects.