I’m constructing an iOS app utilizing React and Capacitor that makes use of face-api.js for facial recognition. The app runs nice in an internet atmosphere each regionally and on prod for each home windows and IOS. Nonetheless, I encounter points when loading the face-api.js fashions on xcode in growth mode. Particularly, I’m getting a tensor form mismatch error that I can not seem to resolve.
Error:
This error happens as quickly because the fashions are loaded, even earlier than I try to make use of them for scanning faces. Right here’s the total error message:
Error loading face-api.js fashions:
Error: Based mostly on the offered form, [1,1,512,9], the tensor ought to have 4608 values however has 1324
at lt — util.ts:108
at $r — tensor_ops.ts:87
at o — io_utils.ts:175
at Yd — io_utils.ts:116
...
at AddEmployeePage.js:38
What I’ve Tried:
1- Mannequin Loading Code:
I’m utilizing the next code to load fashions from a CDN. This works completely on the internet and android apps, however throws the error above on iOS.
const loadModels = async () => {
strive {
const modelUrl = "https://cdn.jsdelivr.internet/gh/justadudewhohacks/[email protected]/weights";
await faceapi.nets.ssdMobilenetv1.loadFromUri(modelUrl);
await faceapi.nets.faceRecognitionNet.loadFromUri(modelUrl);
await faceapi.nets.faceLandmark68Net.loadFromUri(modelUrl);
console.log("Fashions loaded efficiently");
} catch (err) {
console.error("Error loading face-api.js fashions:", err.message);
}
};
2- Loading from Native Recordsdata:
I’ve tried loading fashions from the app’s native filesystem, however the identical challenge persists. I ensured that each one mannequin recordsdata are accurately positioned and accessible.
3- Various Loading Strategies:
I additionally tried to load fashions by way of an offline CDN with no distinction in end result.
Further Info:
Setting:
-
React (with Capacitor for native iOS construct)
-
face-api.js model: 0.22.2
-
Hosted the net model efficiently on Vercel.
Goal:
I would like the facial recognition to operate offline in addition to in net environments. The app might be deployed on a number of iOS gadgets and should deal with real-time facial recognition effectively.
If in case you have any questions for me, please do not hesitate to ask!
Any insights or workarounds could be significantly appreciated!