Skip to content

An Editor to manually adjust automatic Face Mesh Detections

License

Notifications You must be signed in to change notification settings

hcmlab/FaceMeshEditor

Repository files navigation


Face Mesh Editor

An Editor to manually adjust automatic Face Mesh Detections


Demo

Screenshot of the Demo as a Preview

Features

General

Image(s) Upload

Supported formats: jpg and png

Annotation Upload

Supported format: json

Annotation Download

Format: json

History
  • Redo: Allows users to repeat a previous action. Useful for reverting to a specific state.
  • Undo: Reverts the most recent action. Handy for correcting mistakes.
  • Reset: Restores the system to its initial state. Useful for starting over.
Model Selection
  • MediaPipe (Default): Works offline. This mode utilizes a pre-trained model within the system.
  • Webservice (Custom): Operates online. Users can connect to a custom external service for specialized processing.
Delete Features

If a displayed face lacks a trait, marking it as hidden signifies its non-existence, and eliminated points are flagged as such in the json file.

Image Preview

Displays thumbnails of uploaded images. These small visual representations help users quickly identify and select the desired image.

Editor

Cascading Drag

The Cascading Drag tool facilitates smooth movement of points by considering their neighboring points. When a user drags a point, the tool calculates the movement distance for nearby points using an incremental function. This real-time update ensures a responsive and intuitive interface.

Hide Tesselation

The Hide Tesselation feature enhances focus on specific regions during annotation. Even when tesselation and their corresponding points are hidden, users can still manipulate points via cascading drag if they are selected closely enough. This allows precise adjustments while maintaining a clutter-free view.

Zoom

The Zoom functionality allows users to adjust the magnification level of the image. By zooming in, users can examine finer details with clarity. Conversely, zooming out provides a broader view of the entire image.

Pan

The Pan feature enables users to shift the visible portion of the image. By panning, users can explore different areas without changing the zoom level. It’s particularly useful for navigating large or detailed images.

Custom Webservice API

As a user, you have the flexibility to integrate various face mesh detection models seamlessly into our application. Whether you want to provide a pre-trained model via the webservice or create a dynamic model that continues to learn, this API empowers you to tailor face mesh detection to your specific needs.

Key Endpoints

/detect
  • Method: POST and GET
  • Input: Form-data with a file (supported formats: jpg and png)
  • Return: JSON response containing detection results
  • Description: This endpoint allows users to submit an image for detection. The supported formats are jpg and png. The system processes the uploaded file and responds with detection information in JSON format.
/annotations
  • Method: POST
  • Input: JSON string
  • Description: Users can send annotation data in JSON format to this endpoint. The system processes the annotations and performs relevant actions based on the provided information.