You can deploy this model and start earning money today!

jashwanthpeddisetty0712/
ppe-detection-yolov8m

Name: ppe-detection-yolov8m
Availability: InStock

Views

No views yet

PPE Detection — YOLOv8m

Fine-tuned YOLOv8m for construction-site personal protective equipment (PPE) compliance. Detects PPE worn and, critically, PPE violations (missing hardhat, mask, safety vest, gloves, goggles).

Used as a built-in pattern-alert detector in the visual-ai platform; downloaded automatically at runtime via huggingface_hub.hf_hub_download.

Classes (14)

0  Fall-Detected     5  Mask          10  NO-Safety Vest
1  Gloves            6  NO-Gloves     11  Person
2  Goggles           7  NO-Goggles    12  Safety Cone
3  Hardhat           8  NO-Hardhat    13  Safety Vest
4  Ladder            9  NO-Mask

Violation (alert) classes: NO-Hardhat, NO-Mask, NO-Safety Vest, NO-Gloves, NO-Goggles.

Metrics (validation, best checkpoint, 22,077 instances)

Metric	Value
Precision	0.70
Recall	0.86
mAP@50	0.79
mAP@50-95	0.53

Per-class recall / mAP@50 for the violation classes:

Class	Recall	mAP@50
NO-Hardhat	0.93	0.74
NO-Gloves	0.91	0.91
NO-Goggles	0.97	0.96
NO-Mask	0.88	0.66
NO-Safety Vest	0.32	0.21

Note: NO-Safety Vest is the rarest class and has weak recall — treat its alerts with lower confidence and consider oversampling/more data for a future revision.

Training

Base: yolov8m.pt, 640px, batch 16, ~92/100 epochs (early stop, patience 25), seed 0
GPU: RTX 4070 (12 GB), CUDA 12.1

Usage

python

from huggingface_hub import hf_hub_download
from ultralytics import YOLO

path = hf_hub_download(repo_id="jashwanthpeddisetty0712/ppe-detection-yolov8m", filename="best.pt")
model = YOLO(path)
results = model.predict("image.jpg", conf=0.25)