Camera Vision

Offline AI Computer Vision competition. Teams develop Python-based edge algorithms capable of detecting predefined objects in real-time without cloud connectivity.

Mission Briefing

WARNING! EMERGENCY ALERT RECEIVED:

A secure perimeter requires automated surveillance. Your team must develop and code a real-time computer vision system. The script must utilize a live camera feed to instantly detect and classify various objects (such as humans, vehicles, or flags). The system must actively render the correct identification label directly above each detected object on the live monitoring screen.

Competition Overview

Teams deploy machine learning models on their local computers to identify predefined objects in real-time from a webcam feed. The track simulates an offline surveillance station where edge algorithms must detect targets without cloud connectivity.

Mission Objectives

  1. Initialize the local testing hardware.
  2. Process the live video stream offline.
  3. Detect target classes (Humans, Vehicles, Flags, etc.).
  4. Localize targets with accurate bounding boxes.
  5. Display confidence thresholds for each detection.
  6. Log coordinates and metrics automatically.
  7. Complete the evaluation run.

Preliminary Round: Technical Evaluation

Online technical evaluation & layout review. Teams must submit their initial blueprints and project details to our official Gmail address for pre-approval before the main event.

Preliminary Round Registration Requirements

Participants must send a PDF to our official Gmail (aditya.himbotics@gmail.com) containing the following details:

  • School Name
  • Team Name
  • Participant Names
  • Mentor/Teacher Name
  • Contact Details

Technical Evaluation Requirements

In your PDF submission, you must clearly define the following:

  • Project Objective: What specific task is the vision system performing (e.g., security access, object tracking)?
  • Software Stack: What programming languages and libraries (e.g., Python, OpenCV) are you using?
  • Hardware Specifications: What camera and processing unit (e.g., laptop webcam, Raspberry Pi) are you utilizing?
  • System Flowchart: Provide a basic block diagram showing how the system processes an image to achieve the final result.

Track FAQs

Q: Can we use a custom-trained YOLO model?

Yes. Custom weights trained on your own datasets are allowed, provided the inference executes locally and offline on your computer.

Q: Is color masking alone sufficient to pass the tests?

No. The targets are complex physical images representing humans, dogs, cars, flags, and plates. Simple color filter thresholds will fail to classify them accurately.

Q: What camera hardware is supplied at the venue?

Teams bring their own webcams or microcontroller cameras. The venue only supplies standard mounting platforms and local AC power.

Official Participant Handbook (PDF)

READY TO JOIN THE LEAGUE?

Space on the grid is limited to 30 teams per category. Assemble your hardware, verify consents, and register before September 15.

Start Team Registration