Sherlock

Sherlock is a human-centric computer vision library that provides optimized algorithms for hand tracking and face detection in real-time.

Tasks

General

• Generate documentation with Sphinx.
• Deploy API to Amazon AWS Lambda with Kinesis Streams using Boto3.

Infrastructure

• Refactor from functional programming to OOP.
• Add tests and bench mark with sample pictures.

Face Detection

• Implement fast face detection with previous frame data cache.
• Migrate to current infrastructure.
• Modularize functionality.
• Create fast haar cascade module.
• Implement 3D pose estimation with feature tracking and cyclindrical fitting.

Handtracking

• Implement open palm hand tracking.
• Migrate to current infrastructure.
• Modularize functionality.
• Implement advanced methods for hand contour filtering with fast haar cascades.
• Train custom haar cascades for multiple hand positions.
• Implement advanced methods for hand model fitting.

Features

• Handtracking
• 3D Face Pose Estimation
• Object Recognition
• Classifying Human Activity in Videos

License

Apache Software License