Deep Multi-modal Object Detection and Semantic Segmentation for Autonomous Driving: Datasets, Methods, and Challenges

July 2019

tl;dr: Very good review of sensor fusion in 3D object detection.

Overall impression

The summary of data fusion schemes is highly informative. After all, it seems that the pioneering work of MV3D and AVOD are still highly relevant.

Deep learning on Radar data has not been widely explored yet and have potential to yield better object detection algorithms.

Key ideas

Technical details