Cross-Recessed-Screw Deep Learning Datasets

This repository contains the following cross-recessed-screw (CRS) datasets for use with the Tiny-You Only Look Once v2 (Tiny-YOLO v2) deep learning object detection system:

1. Training dataset: 900 images and corresponding ground-truth annotation files of CRS embedded in general hardware.
2. Test Set A: 90 images and corresponding ground-truth annotation files of CRS embedded in general hardware. This set may be used for training validation.
3. Test Set B: 90 images and corresponding ground-truth annotation files of CRS embedded in laptops.
4. Test Set C: 90 images and corresponding ground-truth annotation files of CRS embedded in boxed electronics including power supplies, power tools, and oscilloscopes.

These datasets are located in the separate branches of this repository. Each image is exclusive to its own dataset. All images were taken with a Google Pixel 12.3 MPx (3036 x 4048 pixels) camera.

The research which developed these datasets is available through the following link: https://doi.org/10.1016/j.array.2021.100094.

The authors include Daniel P. Brogan, Nicholas M. DiFilippo, and Musa K. Jouaneh.