Data collection: Set test ranges of distances, angles, and resolutions. Implemented depth camera, QuickSort  algorithm, camera stand marked with angle scale to finish the distance detection, angle measurements, and capture images with various resolutions; Each test set contains 20 to 30 images.

Data preprocessing: Applied image augmentations (Crop, Rotate, Color distortions) to improve the performance of detectors.

Detectors: Pretrained YOLOv4, FasterRCNN, and YOLOv5 on COCO dataset with corresponding pretrained feature extractors on ImageNet dataset.