DetectX PPE

DetectX PPE helps ensure compliance with personal protective equipment (PPE) requirements by detecting hardhats, masks, and vests, as well as identifying missing safety gear in high-risk environments.
The primary purpose is to detect when people are not wearing a hardhat, vest, or mask, so the ACAP also detects NO_Hardhat, NO_Vest, and NO_Mask.

This ACAP application builds on the DetectX base model, trained on 25,000 images across multiple scenarios and lighting conditions, using a custom object detection model derived from several public PPE datasets.

Upon detection, the ACAP can trigger an event, send an MQTT message with bounding box data, or upload a cropped detection image to an HTTP endpoint or MQTT broker.

Example

Download DetectX PPE

Note that the start up time before the ACAP gets responsive may take up to 30-45 seconds (it is loading the model into the memory).

Labels and Detections

Each detection label can be individually toggled on or off:

  • Hardhat
  • Mask
  • Vest
  • NO_Hardhat
  • NO_Mask
  • NO_Vest
  • Person

User & Integration Guide

Refer to DetectX: COCO for usage and integration details.

Model Overview

Dataset: Trained on a number of public PPE datasets Supported Platforms:

Platform Model Input Model Size
ARTPEC-8 960 Small
ARTPEC-9 1440 Small
ARTPEC-9Plus 1440 Medium

Note: ARTPEC-9 provides enhanced inference performance and supports larger model inputs for improved accuracy.

image