AI Door Control

The project involves the development of an AI-based door control system that utilizes computer vision and deep learning technologies to manage boardings and exits through real-time video analysis. This system aims to detect individuals and objects in doorways, replacing traditional sensors with a single AI-powered solution. The project includes the implementation of deep learning models for people and object recognition, movement analysis, real-time processing, decision-making, and logging. The system must process video data on the edge, requiring models optimized for embedded deployment while ensuring detection accuracy and speed. It should function reliably under varying lighting conditions and handle complex scenarios such as multiple passengers, carried objects, and potential obstructions while adhering to strict safety standards. Task Descriptions: Task 1: Data Analysis and Annotation - Categorize data for analysis - Annotate data for training and validation purposes Task 2: Model Evaluation and Performance Analysis - Conduct comprehensive testing to generate detailed daily performance reports - Analyze and document detection failures and performance issues - Develop recommendations for model improvements based on observed failure patterns Task 3: Model Enhancement and Optimization - Prepare data for model training - Develop and optimize deep learning models Task 4: Extended Vision Capabilities - Enhance the system's vision capabilities for improved performance Required Skillset: - Strong expertise in computer vision and deep learning, particularly with PyTorch / TensorFlow and object detection models - Experience in data annotation, preparation, and validation for computer vision applications - Proficiency in model optimization, testing, and deployment on edge devices - Familiarity with MLOps tools such as AWS Sage Maker, DVC, and version control systems Input: - Existing labeled dataset from current deployments - New video data from existing installations - Current model architecture and weights - System performance requirements and specifications,