A 3D mode for a website using webcam and face detection
- NextJs 14
- TensorflowJs
- MediaPipe Face Detection
- HTTPS on local environment: https://mswjs.io/docs/recipes/using-local-https/#trust-certificate-on-browser-level
- Service Worker
- Chrome --> Application Tab --> Service Workers --> Update on reload [checked!]
- Chrome --> Network Tab --> Disable cache [checked!]
- Create a hide & seek website where the user can enable 3D mode using the webcam in order to find hidden things
- A famous 3D painting?
- Alternative point of views:
- Far away (frame?)
- Very close (Keyhole?)
- Recreate famous paintings in 3D
- Enable webcam
- Read webcam stream
- Show video on page
- Once the video is loaded, show "Enable 3D" button
- Enable 3D
- Load the model
- Start face detection
- Read right eye position
- Calculate distance from the center
- Apply translation coords to child elements (Box3D)
- Disable 3D
- Stop face detection
- Stop video
- Position content in the basic position (0,0)
- Disable webcam
- Hide video & Reset Stream
- Enable mouse
- Listen to mouse move events (change x and y position)
- Listen to mouse wheel event (zoom in/out)
- Disable mouse
- Remove listeners (mouse move, mouse wheel)
- Add local mediapipe detector solutionPath
- Cache model using Service Worker (do not download model every time)
- Add React Context in order to simplify the set of x and y point
- Add/Improve depth using both eyes (it works only calculating the horizontal distance)
- Create a more graphical example (Magritte paintings)
- Add threshold/interpolation for face detection (view position buffer)
- Make z (zoom) work not just if eyes are horizontal
- Select webcam (in case of multiple webcams)
- Improve UI
- Improve performance (lighter images?)
- Create an example of a real "standard" website with 3D mode
- TensorflowJs
- MediaPipe/FaceDetection
- Prevent Default inside passive elements
- Emoji Cursor