Msanthosh08/detr-resnet-50-hardhat-finetuned

Helmet Detection using DETR

This model performs object detection on images of construction sites to detect the presence or absence of safety helmets. It is a fine-tuned version of Facebook's DETR model, trained using the hf-vision/hardhat dataset. It is suitable for use in workplace safety monitoring and enforcement in industrial environments.

Model Details

Model Description

This model detects three categories of objects in construction site images:

• With helmet
• Without helmet
• Head It is based on the DEtection TRansformer (DETR) architecture and fine-tuned using Hugging Face Transformers.
• Developed by: Santhosh M
• Model type: Object Detection Transformer (DETR)
• Language(s): English
• License: Apache 2.0
• Fine-tuned from model: facebook/detr-resnet-50

Model Sources

• Repository: Msanthosh08/detr-resnet-50-hardhat-finetuned
• Paper: DETR: End-to-End Object Detection with Transformers

Uses

Direct Use

This model can be used directly to:

• Monitor workers on construction sites in real-time or from recorded video
• Detect safety violations like workers without helmets
• Aid in automated safety compliance reporting

Bias, Risks, and Limitations

• May fail in poor lighting or occluded views
• Might not generalize to unseen environments or new helmet types
• Dataset may carry biases (e.g., region, worker demographics)

Recommendations

Users should validate the model on their specific deployment scenarios. Do not rely solely on this model for safety-critical decisions without human supervision.

Model Card Authors

• Santhosh M

Contact

• [email protected]