The dataset viewer is not available because its heuristics could not detect any supported data files. You can try uploading some data files, or configuring the data files location manually.

Indian Traffic Helmet Violation Detection (Deduplicated & Rebalanced)

Part of the Edge-AI Traffic & Vehicle Analytics System repository by thundarstrom.

Dataset Summary

A unified, audited, and strictly deduplicated dataset of 42,559 images and ~126,000 bounding boxes for two-wheeler helmet compliance monitoring on Indian roads.

Key Features

Zero Train/Test Leakage: Fully verified perceptual hash partition (0 matching image hashes across splits).
Balanced Evaluation Split: 80/10/10 split ensuring representative test benchmarking.
Classes:
- 0: helmet (Full-face, open-face, construction/half helmets)
- 1: no_helmet (Bare heads, turbans, caps, scarves without safety helmets)

Splits

Train: 34,047 images (80%)
Validation: 4,255 images (10%)
Test: 4,257 images (10%)

How to Access and Download

Using Automated Project Downloader (Extracts automatically)

# Clone / pull and auto-extract dataset
python scripts/download_hf_datasets.py --dataset helmet --org thundarstrom

Using `huggingface_hub` Python SDK

from huggingface_hub import snapshot_download

# Download into local dataset directory
local_path = snapshot_download(
    repo_id="thundarstrom/traffic-helmet-violation",
    repo_type="dataset",
    local_dir="data/datasets/helmet_combined"
)
print(f"Dataset downloaded to: {local_path}")

Recommended Training / Evaluation Recipe

python scripts/train_models.py --task helmet --model yolov8s.pt --epochs 120 --imgsz 640 --batch 32 --fl_gamma 1.5

Citation & Maintainer

Maintained by: thundarstrom
Project: Edge-AI Real-Time Traffic Violation Detection & ANPR
License: CC-BY-4.0

Downloads last month: 49

Total file size:

2.63 kB