Open jojokic opened 1 week ago
regarding your issue with the frontend models, MediaPipe (the library I use) has external dependencies:
https://cdn.jsdelivr.net/npm/@mediapipe/tasks-vision@0.10.3/wasm
https://storage.googleapis.com/mediapipe-models/face_landmarker/face_landmarker/float16/1/face_landmarker.task
I suppose a branch of FacePoke could be created, one that would put those assets inside the repo itself, and version them using git lfs add "*.task"
or git lfs add "*.wasm"
(they aren't big, the model is less than 5 Mb I think)
I'm not sure about how it would work for the first link (the one with /wasm), it seems to be valid (it comes from the official code examples and README) but when we open this link it leads to nowhere, so that's weird. If someone has an explanation for that and can demonstrate the asset can be versioned easily in the repo itself, then we could add it.
(note that the backend will still have to connect to the web ie to Hugging Face to fetch some models, although normally this is a one time operation)
Please I need your help! Did you find any way to run it completely offline?
Hello, when I disconnect the Internet, it will appear:
“error Failed to initialize face detection. Please try refreshing the page.”
What kind of question is that? How can I disconnect from the Internet and using this code? thank you.