LiDAR-Fusion Chapter 4 - Detect Objects in ZTE Dataset

In this chapter, we will utilize the trained model to detect objects in the ZTE dataset using PointPillars. The result will be generated in .txt format for later tracking.

Detect Objects

Since we have already obtained a trained model from the last step, we can directly use it to do the detection. You can simply run detect.py in the detector/tools directory. Everything will be handled automatically.

You can refer to parser_config() at the beginning of the script and modify some parameters if you want to change a checkpoint or data path.

Result Explanation

The detection result will be stored in the detector/tools/detection_result. It contains a log file and three result files.

The three result files refer to:

• result_1.txt - Detection results of cars.
• result_2.txt - Detection results of pedestrians.
• result_3.txt - Detection results of cyclists.