I have started learning about 3D object detection and segmentation (instance also). So I thought to make a list of all papers I am reading
PointNets and PointNets++
PointNets started it at the beginning (earlier 3D data was not directly used, but PointNets used networks directly on depth. paper1paper2. I would mostly skim through these.
Frustum PointNets
Extended PointNets, by using 2D image and then converting it into a frustum. paper
PointRCNN, IPOD and PointPillars
I have yet to read these, but I selected them as they were at rank 1, 2, 3 in various categories of KITTI vision dataset. paper1paper2paper3. Also, these papers were published in December 2018 on arxiv so they are pretty new.
This dataset aims to democratize access to such data, and foster innovation in higher-level autonomy functions for everyone, everywhere. By conducting a competition, we hope to encourage the research community to focus on hard problems in this space—namely, 3D object detection over semantic maps.
In this competition, you will build and optimize algorithms based on a large-scale dataset. This dataset features the raw sensor camera inputs as perceived by a fleet of multiple, high-end, autonomous vehicles in a restricted geographic area.
I never worked with 3D object detection before. But based on my experience with Retina net and SSD, they takes longer to train and debug. With this new 30 hrs / 7 days rule on Kaggle, I really don’t know if I got the resource to play with it
But I would love to hear how people solve this problem