Better photo detection

[Madnex]

This pull request introduces support for multiple photo detection modes in ScanSplitter, including a new deep learning-based AI (U2-Net) mode, and adds robust handling and UI feedback for downloadable models (U2-Net and orientation). The changes span user documentation, core frontend logic, settings UI, API calls, and type definitions, enabling users to choose detection modes, manage model downloads, and see real-time status updates.

Detection Modes & Model Management

• Added support for three detection modes: ScanSplitterv1, ScanSplitterv2 (default), and AI (U2-Net); users can now select the mode in the UI, and the backend is called with the appropriate parameters. [1] [2] [3] [4] [5]
• Introduced model download management: the frontend checks model status and triggers downloads as needed for U2-Net and orientation models, polling for progress and handling errors. [1] [2] [3] [4] [5] [6] [7]

Frontend UI Enhancements

• Updated the settings panel to allow detection mode selection, toggle between U2-Net lite/full models, and display real-time download progress or errors for required models. [1] [2] [3] [4]
• Passed model status information from App to the SettingsPanel for user feedback. [1] [2]

API and Types

• Extended API functions and types to support detection mode and model download operations, including new endpoints for model status and download, and new types for detection modes and model statuses. [1] [2] [3] [4] [5] [6]

Documentation

• Updated the README.md to document the new detection modes, model download requirements, and updated CLI usage and options. [1] [2] [3]

These changes collectively improve ScanSplitter's flexibility, user experience, and robustness when handling advanced detection scenarios and large models.