PPE Detector for Worker Safety

The PPE Detector for Worker Safety - is a real-time computer vision model for PPE non-compliance detection in the industrial setting. The solution detects seven object classes: Bare Head, Helmet, Ear Protection, Welding Mask, Bare Chest(NO Visibility Vest), High Visibility Vest, Person.

Usage Information

Using our model for real time prediction is as simple as this:

predictor = sagemaker.predictor.RealTimePredictor(
    ' your endpoint name ',
    sagemaker_session=sagemaker.Session(),
    content_type="image/jpeg"
)

with open('data/sample_image.jpg', 'rb') as img:
    img_bytes = bytearray(img.read())
    result = predictor.predict(img_bytes).decode("utf-8")

Also we've published two notebooks showing how to use our model:

• Using-PPE-Detector-Endpoint.ipynb notebook shows how you can use Python API to perform inference on endpoint created from the model
• Using-Personal-Protection-Equipment-Detection-model.ipynb notebook shows how you can use Python API to run the full scenario:
  • deploy our model to create an endpoint
  • run Real Time inference on endpoint using local image
  • visualize and save the prediction on original image
  • run Batch Transform job to perfom the inference on your data stored in Amazon S3 bucket

Input\output data examples

• You can find sample input data in demo_input folder
• demo_output_images folder contains images with detections predicted by the model and visualized using utils.visualize_detection method
• demo_raw_output folder contains raw output generated by our model using Batch Transform approach

Sample detection results: