Opportunity brief
About the Event:
VISION QUEST is a beginner-friendly, hands-on Computer Vision Challenge where participants explore how AI interprets visual data to solve real-world challenges. Combining image processing, object detection, classification, and visual analysis across four competitive rounds, the event is aligned with key UN Sustainable Development Goals:
- SDG 3: Good Health & Well-being (safety, accessibility, and visual health analysis)
- SDG 11: Sustainable Cities & Communities (traffic, roads, public spaces, and urban infrastructure)
- SDG 15: Life on Land (monitoring plants, wildlife, and environmental elements)
Whether using provided starter kits or building custom models from scratch, teams will test their technical problem-solving, model application, adaptability, and real-world vision capabilities.
Round Wise Details:
Round 1: Can You See It?
- Format: Fast-paced visual quiz & observation challenge.
- Mechanism: Teams answer image-based questions testing foundational Computer Vision concepts, including object identification, image transformations, color spaces, patterns, and visual reasoning.
Round 2: Teach Your AI
- Format: SDG-themed hands-on development sprint.
- Mechanism: Teams receive an SDG-inspired dataset and problem statement. Using Python, OpenCV, pre-trained models, or Vision APIs, participants build a functional vision pipeline. Starter resources are provided for beginners.
Round 3: Surprise Vision
- Format: Generalization & robustness test.
- Mechanism: Solutions are subjected to a brand-new, unseen evaluation dataset featuring unfamiliar scenarios and environmental variations to test how well the models generalize.
Final Round: See It. Spot It. Solve It.
- Format: Live demonstration & technical defense.
- Mechanism: Top-performing teams receive a final vision challenge, demonstrate their pipeline in real time, explain their methodology, and detail how their approach applies to the chosen SDG context.
Rules and Guidelines:
- Team Composition: Teams must consist of 2–4 participants.
- Tech Stack & Tools:
- Primary focus must be on Computer Vision techniques (OpenCV, NumPy, Pandas, PyTorch, TensorFlow, pre-trained vision models, or permitted Vision APIs).
- Participants are not required to train models from scratch; pre-trained weights and APIs are allowed unless explicitly restricted for a round.
- External datasets or pre-built, complete solutions are strictly prohibited unless permitted by the organizers.
- Submission & Deadlines: Strict time limits apply for each round. Late submissions will face point penalties or disqualification.
- Evaluation Metrics: Solutions will be judged using standard Computer Vision metrics (Accuracy, Precision, Recall, F1-Score, IoU, mAP) alongside challenge-specific criteria.
- Academic Integrity: Plagiarism, copying code between teams, unauthorized external help, or code manipulation will result in immediate disqualification.
- Authority: Organizers reserve the right to modify round structures or datasets if required. The judges' decisions on scoring and qualifications are final and binding.