Integrating ADAS with Keypoint Feature Pyramid Network for 3D LiDAR Object Detection

This repository contains the training scripts for Keypoint Feature Pyramid Network, specifically for 3D LiDAR Object Detection. In this case, the KITTI 360 Vision dataset has been used to train the detection model.

It is part of the LearnOpenCV blog post - Integrating ADAS with Keypoint Feature Pyramid Network for 3D LiDAR Object Detection.

Environment Setup

Run the following commands on a new terminal window for creating a new environment with the required packages:

cd SFA3D
pip install -r requirements.txt

Dataset Visualization

To visualize 3D point clouds with 3-dimensional bounding boxes, run the following commends:

cd sfa/data_process
python kitti_dataset.py

Inference

There is an instance of a pre-trained model in this repository. You can use it to run inference:

python test.py --gpu_idx 0 --peak_thresh 0.2

Video Demonstration

Similarly, inference can be run on a video stream:

python demo_2_sides.py --gpu_idx 0 --peak_thresh 0.2

Training Pipeline

Single Machine w/ Single GPU

python train.py --gpu_idx 0

Single Machine w/ Multiple GPUs

python train.py --multiprocessing-distributed --world-size 1 --rank 0 --batch_size 64 --num_workers 8

Evaluation Metrics - TensorBoard

To track the training progress, go to logs/ folder and run:

cd logs/<saved_fn>/tensorboard/
tensorboard --logdir=./

Then, just go to http://localhost:6006/

AI Courses by OpenCV

Want to become an expert in AI? AI Courses by OpenCV is a great place to start.