Competition

Road Sign Detection

Organized by ayoubsa - Current server time: March 30, 2025, 7:16 a.m. UTC

Current

Development
Nov. 4, 2024, midnight UTC

End

Competition Ends
Dec. 19, 2025, midnight UTC

Road safety is one of the most crucial challenges in modern transportation. Autonomous vehicles must accurately detect and classify road signs to follow traffic rules and navigate safely. Misinterpretation or missed recognition of road signs, such as speed limits or stop signs, can result in accidents or traffic violations. Thus, developing a highly accurate detection system is essential for enabling self-driving cars to operate safely in real-world conditions.

The goal of this challenge is to build a multiclass object detection model that can identify and localize various types of road signs in diverse traffic scenarios. Participants are required to develop a machine learning model capable of detecting traffic lights, speed limits, and stop signs—key indicators for safe and lawful driving.

The development of such models will significantly improve the reliability of autonomous vehicle systems, ensuring that vehicles comply with traffic regulations in complex environments. An efficient detection model can be used in large-scale deployment for self-driving fleets, reducing the dependency on human drivers and minimizing traffic incidents caused by human error.

The error metric for this competition is accuracy.

Your submission file should look like this:

image id‎ ‎ ‎ ‎ ‎ ‎ ‎ ‎ class‎ ‎ ‎ ‎ ‎ ‎ ‎confidence‎ x_center‎ y_center‎ ‎ width‎ ‎ ‎ height

ID_2TZLLT80‎ ‎ ‎ ‎ ‎ ‎ 5‎ ‎ ‎ ‎ ‎ ‎ ‎ ‎ ‎ ‎ ‎ ‎ ‎ ‎ 0.5‎ ‎ ‎ ‎ ‎ ‎ ‎ ‎ ‎ ‎ ‎ 130‎ ‎ ‎ ‎ ‎‎ ‎ ‎ ‎ ‎ ‎ ‎ 12‎ ‎‎ ‎ ‎ ‎ ‎‎ ‎ ‎ 340‎ ‎ ‎ ‎‎ ‎ ‎ ‎‎ ‎ ‎ ‎300

  • Image_Id: is the Id assigned to each image. Note, that each image can have more the one object , which translates to more than one bounding box.
  • Class: is the particular bounding box classification, i.e.,, Green light, stop,...
  • Confidence score: each object detector model gives the confidence score of each bounding box predicted in an image; this value is used to sort the bounding boxes.
  • (x_center, y_center, width, height): The values of the bounding box.

Please note that we will not tell you if you are missing an image in your submission file. You will need to make sure you have submitted a prediction for each image.

The dataset contains images of road signs with annotations in YOLO format, which specify the class ID and the bounding box coordinates for each object.

There are 15 classes:

  • Traffic Lights: Green Light, Red Light
  • Speed Limits: 10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 110, 120
  • Stop Sign: Stop

Each image can contain multiple road signs. The dataset simulates real-world driving conditions, including varying weather, lighting, and road environments.

The dataset is splitted into train, validation and test. Here are the links for each one:

Development

Start: Nov. 4, 2024, midnight

Description: Submit your submission.zip with a predictions.csv file in YOLO format.

Competition Ends

Dec. 19, 2025, midnight

You must be logged in to participate in competitions.

Sign In
# Username Score
1 ayoubsa 0.4980