feat: Implement Flask-based web interface for Deep-Live-Cam
This commit introduces a significant architectural change, transforming the
Deep-Live-Cam application from a desktop-focused customtkinter application
to a web-based interface powered by Flask.
Key changes and features implemented:
1. **Web Framework & UI Basics:**
* Integrated Flask (`webapp.py`) as the backend framework.
* Added basic HTML (`templates/index.html`), CSS (`static/style.css`),
and JavaScript (`static/main.js`) for the frontend structure.
2. **Core Logic Refactoring:**
* `modules/core.py` was refactored to decouple it from UI-specific
code, allowing its functions to be used as a library by both the
new web backend and the existing CLI (`run.py`).
* Utility functions in `modules/utilities.py` were updated for
better modularity and explicit path handling.
3. **Web UI Functionality:**
* **File Uploads:** Implemented API endpoints and frontend JavaScript
to allow you to upload source face images and target media
(images/videos). Previews are shown for selected files.
* **Settings Management:** You can toggle processing options (e.g.,
keep FPS, mouth mask) in the web UI, and these settings are sent
to the backend to update global configurations.
* **File Processing:** A `/start_processing` API endpoint triggers the
core face swapping logic on uploaded files. Processed files can be
downloaded via a separate endpoint. (Current processing is synchronous).
* **Live Webcam Preview:**
* A `/video_feed` endpoint streams live camera input to the browser.
* This feed now incorporates real-time face swapping and other
selected processing effects (e.g., FPS counter, enhancers if toggled),
using the uploaded source face and current settings.
* **Face Mapping via Web UI:**
* Implemented functionality for you to define specific source faces
for multiple target faces detected in an uploaded image.
* Backend APIs were created to analyze target images, send detected
face data (as base64 images) to the frontend, and then receive
your submitted mappings (source images per target face).
* The `face_swapper.py` module was enhanced with `v2` processing
methods (`process_image_v2`, `process_video_v2`, updated
`process_frame_v2`) that utilize these detailed mappings.
* Both file-based processing and the live webcam feed now respect
these mappings if the "Map Faces" option is enabled.
This lays the groundwork for a fully featured web version of Deep-Live-Cam.
Further work will focus on UX refinements (like async processing, better
error handling, UI for clearing mappings) and comprehensive testing.
pull/1375/head
google-labs-jules[bot]
parent
6210b067cc
commit
b7f92acb34