DEV Community

Praveen Amancharla
Praveen Amancharla

Posted on

Indian Road Sign Detection — an API Built for Indian Roads

Most traffic sign recognition systems are trained on European or US sign datasets. They don't work on Indian roads.

India follows its own standard — IRC:67-2012 from the Indian Roads Congress — with over 100 sign types across mandatory, prohibitory, warning, informatory, and hazard marker categories. The shapes, colors, and symbols are different from Western standards.

On top of that, real-world conditions make detection harder: faded paint, dust, non-standard placement, night glare, rain.

I built an API that detects 78 of these IRC sign classes from dashcam imagery.

What It Does

Upload a dashcam frame, get back detected signs with type, confidence, and bounding box coordinates.

  • 78 IRC sign classes — mandatory, prohibitory, warning, speed limits, hazard markers, informatory
  • 96.7% mAP50 accuracy on real Indian dashcam footage
  • ~0.2 second inference per image
  • Speed limit values — detects the specific speed (40, 60, 80 km/h)
  • High precision — tuned to minimise false positives in real Indian road conditions ⸻ The 78 Classes Category Count Examples Warning 34 go_slow, school_ahead, pedestrian_crossing, falling_rocks Prohibitory 14 no_entry, no_parking, one_way, overtaking_prohibited Speed Limit 9 20, 30, 35, 40, 50, 60, 80, 100, 120 km/h Hazard Markers 8 chevron boards, object hazard, two-way hazard Mandatory 7 keep_left, keep_right, roundabout, sound_horn Informatory 5 free_left, give_way, stop, height_limit

Full class list: GitHub repo

Quick Start

import requests

url = "https://indian-road-sign-detection.p.rapidapi.com/api/v1/detect"
headers = {
"X-RapidAPI-Key": "YOUR_API_KEY",
"X-RapidAPI-Host": "indian-road-sign-detection.p.rapidapi.com",
}

with open("dashcam_frame.jpg", "rb") as f:
response = requests.post(url, headers=headers, files={"image": f})

for sign in response.json()["detections"]:
print(f"{sign['sign_type']} — {sign['confidence']:.0%}")

Response:
{
"detections": [
{
"sign_type": "maximum_speed_limit_40",
"category": "speed_limit",
"confidence": 0.98,
"bbox": [412, 85, 498, 192],
"ocr_value": "40"
}
],
"inference_time_ms": 187
}


Who Is This For?

  • ADAS / autonomous driving teams — speed limit enforcement, sign-aware navigation for Indian roads
  • Fleet management companies — automated compliance monitoring across routes
  • Insurance telematics — risk scoring from dashcam footage
  • Road infrastructure mapping — sign inventory from street-view imagery
  • Driving schools and assessment platforms — automated evaluation

    Pricing
    Tier Calls/month Price
    Basic 100 Free
    Pro 5,000 $12/mo
    Ultra 50,000 $49/mo
    Enterprise Unlimited Contact us

    Links

  • Try the API: RapidAPI — Indian Road Sign Detection

  • Live demo: skotantrx.com/demo.html

  • GitHub (docs + samples): github.com/amanchap/indian-road-sign-detection

    Built by SKO TantrX — Break the Old. Build the Bold.

Tags: #ComputerVision #AI #DeepLearning #ADAS #IndianRoads #RoadSafety #API #MadeInIndia

Top comments (0)