Filtering tuna cans with defects using computer vision


Project scope
Categories
Machine learning Artificial intelligence Data scienceSkills
anomaly detection python (programming language) computer vision researchThe main goal is to apply computer vision techniques to develop an algorithm capable of processing a large dataset of tuna can images and identifying a minority subset that may represent anomalies or defects. The focus is solely on filtering these potential defect images from the dataset.
Students will need to:
1. Understand the Dataset:
- Analyze the provided dataset of tuna can images.
- Review sample images illustrating the appearance of defects for reference.
2. Develop an Anomaly Detection Algorithm:
- Research computer vision techniques suitable for identifying anomalies (e.g., feature-based anomaly detection or clustering methods).
- Build, test, and refine the algorithm to ensure it effectively identifies potential defective cans.
3. Deliverables:
- A Python script or notebook that takes a large dataset as input and returns a filtered subset of potential defect images.
- Documentation detailing the approach, methodology, and suggestions for improvement.
Sharing knowledge in specific technical skills, techniques, methodologies required for the project.
Providing access to necessary tools, software, and resources required for project completion.
Scheduled check-ins to discuss progress, address challenges, and provide feedback.
About the Community Partner
ThisFish Inc. provides software and artificial intelligence for seafood traceability and production workflows that improves business efficiency and increases trust and transparency in seafood supply chains. Our mission is to improve the social, environmental, and financial sustainability of seafood enterprises by making seafood supply chains more transparent.
Portals

