ashleve / sim2real_object_recognition

Experimenting with synthetic data generation for object detection using photogrammetry and Unit Perception Tools.

Sim2real Object Recognition

Scan real life objects with photogrammetry -> Generate synthetic dataset in simulation using Unity Perception Tools -> train with tiny YOLO -> port to Arduino based vehicle with Nvidia Jetson

Photogrammetry

Software used:

Meshroom (https://github.com/alicevision/meshroom)

Synthetic dataset generation

Software used:

Unity Perception Tools (https://github.com/Unity-Technologies/com.unity.perception)

YOLO

Model used:

YOLOv3 tiny (https://github.com/ultralytics/yolov5)

Prediction examples