Face-api.js: TinyYolov2 - load model before inference

[swapanil]

I am getting this error in production build in React Js and working fine when use npm start.

import * as faceapi from 'face-api.js'

export async function loadModels() {
    const MODEL_URL = process.env.PUBLIC_URL + '/models'
    await faceapi.loadTinyFaceDetectorModel(MODEL_URL)
    await faceapi.loadFaceLandmarkTinyModel(MODEL_URL)
    await faceapi.loadFaceRecognitionModel(MODEL_URL)
}

export async function getFullFaceDescription(blob, inputSize = 512) {
    const scoreThreshold = 0.5
    const OPTION = new faceapi.TinyFaceDetectorOptions({
        inputSize,
        scoreThreshold
    })
    const useTinyModel = true
    const img = await faceapi.fetchImage(blob)
    const fullDesc = await faceapi
        .detectAllFaces(img, OPTION)
        .withFaceLandmarks(useTinyModel)
        .withFaceDescriptors()
}

Please help me on this.

[flatsiedatsie]

I saw the same error in a situation where I wasn't even using the Yolo model, which was confusing.

In the end it was because I had made a coding error, and the models I did want to use weren't actually loaded. Perhaps there is something in the FaceApiJs code where it throws a wrong error for a missing model, and ends up outputting the last name in the list of nets.

[buraksekili]

Hi @swapanil ,
I am faced with same issue. We have same code block. Have you solve your problem?

[buraksekili]

Hey @swapanil ,
FYI, I solved my problem. I couldn't load models correctly. Can u check your models location? It should be under the public folder as follows; public/models

[amit0shakya]

I made it sure my all models are loading correctly infact my TinyYolov2 json and shard file loading correctly but still it showing me same issue, I replace it with SsdMobilenetv1Options it is working for now. I think this should be a bug.

[RezaRahmati]

I added this and issue resolved

faceapi.nets.tinyFaceDetector.loadFromUri(this.MODEL_URL),