EDGE AI / INDUSTRIAL SAFETY

Makerspace Sentinel

Real-time PPE detection, from model training to a live edge deployment at Texas State.

Context: January 2025–Present · Ingram School of Engineering, Texas State University

01

The engineering

Built an end-to-end pipeline from dataset selection to a CBAM-enhanced YOLOv8 model fine-tuned on three PPE classes.
Validated on 18,000+ unseen images, including work with the SH17 and R2PPE benchmarks.
Optimized inference with TensorRT and deployed on NVIDIA Jetson AGX Orin.

02

The outcome

Running live in Ingram with 25–30 ms inference. Supported by a $3,000 Makerspace Digital Twin grant.