PointPillars

Model description

A Simple PointPillars PyTorch Implenmentation for 3D Lidar(KITTI) Detection.

• It can be run without installing mmcv, Spconv, mmdet or mmdet3d.
• Only one detection network (PointPillars) was implemented in this repo, so the code may be more easy to read.
• Sincere thanks for the great open-souce architectures mmcv, mmdet and mmdet3d, which helps me to learn 3D detetion and implement this repo.

Detection Visualization

[Compile]

cd ops
python3 setup.py develop

[Datasets]

1. Download

   Download point cloud(29GB), images(12 GB), calibration files(16 MB)和labels(5 MB)。 Format the datasets as follows:

   kitti
       |- ImageSets
           |- train.txt
           |- val.txt
           |- test.txt
           |- trainval.txt
       |- training
           |- calib (#7481 .txt)
           |- image_2 (#7481 .png)
           |- label_2 (#7481 .txt)
           |- velodyne (#7481 .bin)
       |- testing
           |- calib (#7518 .txt)
           |- image_2 (#7518 .png)
           |- velodyne (#7418 .bin)

   The train.txt、val.txt、test.txt and trainval.txt you can get from:

   wget https://raw.githubusercontent.com/traveller59/second.pytorch/master/second/data/ImageSets/test.txt
   wget https://raw.githubusercontent.com/traveller59/second.pytorch/master/second/data/ImageSets/train.txt
   wget https://raw.githubusercontent.com/traveller59/second.pytorch/master/second/data/ImageSets/val.txt
   wget https://raw.githubusercontent.com/traveller59/second.pytorch/master/second/data/ImageSets/trainval.txt

2. Pre-process KITTI datasets First

   ln -s path/to/kitti/ImageSets ./dataset
   python3 pre_process_kitti.py --data_root your_path_to_kitti

   Now, we have datasets as follows:

   kitti
       |- training
           |- calib (#7481 .txt)
           |- image_2 (#7481 .png)
           |- label_2 (#7481 .txt)
           |- velodyne (#7481 .bin)
           |- velodyne_reduced (#7481 .bin)
       |- testing
           |- calib (#7518 .txt)
           |- image_2 (#7518 .png)
           |- velodyne (#7518 .bin)
           |- velodyne_reduced (#7518 .bin)
       |- kitti_gt_database (# 19700 .bin)
       |- kitti_infos_train.pkl
       |- kitti_infos_val.pkl
       |- kitti_infos_trainval.pkl
       |- kitti_infos_test.pkl
       |- kitti_dbinfos_train.pkl
   

[Training]

Single GPU training

python3 train.py --data_root your_path_to_kitti

Multiple GPU training

python3 -m torch.distributed.launch --nproc_per_node 8 train_dist.py --data_root your_path_to_kitti

[Evaluation]

python3 evaluate.py --ckpt pretrained/your_weights.pth --data_root your_path_to_kitti 

[Test]

# 1. infer and visualize point cloud detection
python3 test.py --ckpt pretrained/your_weights.pth --pc_path your_pc_path 

# 2. infer and visualize point cloud detection and gound truth.
python3 test.py --ckpt pretrained/your_weights.pth --pc_path your_pc_path --calib_path your_calib_path  --gt_path your_gt_path

# 3. infer and visualize point cloud & image detection
python3 test.py --ckpt pretrained/your_weights.pth --pc_path your_pc_path --calib_path your_calib_path --img_path your_img_path


e.g. [infer on val set 000134]

python3 test.py --ckpt pretrained/your_weights.pth --pc_path /home/lifa/data/KITTI/training/velodyne_reduced/000134.bin

or

python3 test.py --ckpt pretrained/your_weights.pth --pc_path data/kitti/training/velodyne_reduced/000134.bin --calib_path data/kitti/training/calib/000134.txt --img_path data/kitti/training/image_2/000134.png --gt_path data/kitti/training/label_2/000134.txt

Acknowledements

Fast Encoders for Object Detection from Point Clouds](https://arxiv.org/abs/1812.05784)