Building datasets for indoor lidar and AR/VR artificial intelligence training

Date
Sep. 2020 - Feb. 2021

Project Objective

To build a dataset for training an object recognition deep learning model to track the continuous movement of pedestrians in a shopping mall using a camera/lidar fusion sensor.
Build a complete basic pipeline of dataset collection, refinement, processing, and training for camera/lidar training
Utilization Services
Identify customer movement information through pedestrian recognition/tracking technology and utilize it for store design, product display strategy, merchandising, marketing, etc.
AI-based 3D lidar security application

Contributions

Identified and arranged locations for indoor pedestrian data collection for AI training
Stayed at the shooting location and collected data using the data collection system
3D lidar point cloud data labeling
Research, experiment, analyze, and evaluate deep learning-based 3D lidar object detection algorithms
Writing user manuals for training deep learning-based artificial intelligence models

References

AIHub - Indoor Lidar and Camera Synchronization Video Data : https://aihub.or.kr/aidata/27742
Introduction Video