search

maudzung / SFA3D

Super Fast and Accurate 3D Object Detection based on 3D LiDAR Point Clouds (The PyTorch implementation)
https://github.com/maudzung/Super-Fast-Accurate-3D-Object-Detection
MIT License
1k stars 271 forks source link

Model trained for larger distance not working #66

Open bhaskar-anand-iith opened 2 years ago

bhaskar-anand-iith commented 2 years ago

Hi,

I tried training the model to detect for a larger distance. I changed the kitti_config.py as follows:

boundary = { "minX": 0, "maxX": 100, "minY": -25, "maxY": 25, "minZ": -2.73, "maxZ": 1.27 }

bound_size_x = boundary['maxX'] - boundary['minX'] bound_size_y = boundary['maxY'] - boundary['minY'] bound_size_z = boundary['maxZ'] - boundary['minZ']

boundary_back = { "minX": -50, "maxX": 0, "minY": -25, "maxY": 25, "minZ": -2.73, "maxZ": 1.27

X-range is increased to 100m. but with this trained model detection is too poor. Is there any mistake my side? What is boundary_back? What are BEV_WIDTH, BEV_HEIGHT and DISCRETIZATION? Do we need to change these also before training for a larger distance?