How laoode/eduview works
Main Components
• Frontend/UI:
• The “object_cheating/components” directory includes modules like camera_feed, coordinate_panel, detection_display, stats_panel, and controls. These represent the dynamic panels and control widgets that make up the user interface.
• The “assets” and “images” directories supply static files (CSS animations, images for the app view, etc.) that will be referenced in the UI.
• Backend/Processing:
• The core functionality is contained within the “object_cheating/object_cheating.py” file and related state management (e.g., “object_cheating/states/camera_state.py”). These handle the logic of real-time camera monitoring and state handling.
• The “object_cheating/utils” (with modules like eye_tracker.py) support specific analytical functions for tracking eye movement.
• AI & Detection Engines:
• Within “object_cheating/models” you see model files (e.g., YOLO and custom CNN models, plus a Mediapipe-related face_landmarker) which are used for object detection, facial recognition, and head movement analysis.
• Data Sources:
• The “uploaded_files” directory contains sample video files that serve as inputs or test cases for detection.
• Configuration & Integration:
• Files like “rxconfig.py” and the “requirements.txt” define the configuration and dependencies for the Reflex (full-stack) framework that integrates the frontend, backend, and machine learning modules.
Relationships and Interactions
• Data Flow:
• Video feeds (from live cameras or uploaded exam footage) are captured via the frontend components (camera_feed) and then sent to the backend processing.
• The backend (object_cheating.py) orchestrates calls to the AI models (YOLO, CNN, Mediapipe) and uses utility modules (like eye_tracker.py) to analyze video frames for suspicious activity.
• The detection results are then fed back into the UI panels (detection_display, stats_panel, table) to update teachers/examiners in real-time.
• Component Interactions:
• Controllers (controls.py) likely handle user inputs and state transitions (perhaps utilizing the “states/camera_state.py”).
• The design likely follows a modular structure where each component deals with a specific part of the exam monitoring process, tied together by the Reflex framework that handles routing, state, and UI rendering.
Architectural Patterns and Design Principles
• Separation of Concerns:
• There’s a clear divide between the presentation layer (UI components), the processing/logic layer (object_cheating core, state tracking), and the AI/detection layer (models).
• Modularity:
• The project is broken down into components, models, states, and utilities, making it easier to update or extend individual parts without affecting the whole system.
• Real-Time Processing & Feedback:
• The system emphasizes low-latency interactions between video input and UI feedback, essential for real-time cheating detection.
Technologies and Frameworks
• Reflex Framework: Serves as the glue between frontend and backend.
• YOLO: Used for object detection.
• Mediapipe: Used for facial landmark detection.
• Convolutional Neural Networks (CNN): Provide additional deep learning capabilities for analysis.
• Python (and its standard packaging structure) ties everything together.
Explicit Elements
• Each major component should be clearly labeled:
• “Frontend Components” (listing panels, controls, camera feed, etc.)
• “Backend Processing” (core logic, state management)
• “AI Models and Detection Engines” (models directory items)
• “Static Assets & Data Sources” (assets, images, uploaded_files)
• Flow from video input → UI capture → backend processing → AI analysis → feedback display.
• Visual distinction: