VoteNet

This is an unofficial implementation of "Deep Hough Voting for 3D Object Detection in Point Clouds" (https://arxiv.org/abs/1904.09664)

Training

• Prepare SUN RGB-D dataset following the instructions of https://github.com/charlesq34/frustum-pointnets/tree/master/sunrgbd, note you only need to run extract_rgbd_data.m
• Compile custom Tensorflow ops as described in PointNet++
• Pass the root folder of generated data in MyDataFlow and run run.py.

TODOs

• Data augmentation
• Test/Validation mAP
• Train/Validation split
• 3D NMS

Results (To be updated)

• After 75 epochs, I got the following AP:

  • table:0.002599436474719265

  • bed:0.3710421555617831

  • night_stand:0.0067538466938640174

  • bookshelf:0.15430493020623803

  • chair:0.05431061678741693

  • dresser:0.035260119681139616

  • sofa:0.12197371148148911

  • desk:0.0022836462245173833

  • toilet:0.010896393557409

  • bathtub:0.0

  • mAP:0.075942