Error of "no faces detected for label" with 400 face images

[larui529]

Hi

I am trying to use tiny face model to do a face recognition api with 400 different faces. But I got "no faces detected for labels" error when I chose any input size larger than 128. The accuracy of the model is not very good based on the small input size. I am wondering if there is any solution to that. Thanks.

Rui

[justadudewhohacks]

"no faces detected for labels " where does this Error come from?

Please share some example images, so I can verify the issue. Also you can try out the SSD mobilenetv1 model for face detection.

[slash-dev]

I experienced a similar problem following the example at https://itnext.io/face-api-js-javascript-api-for-face-recognition-in-the-browser-with-tensorflow-js-bcc2a6c4cf07

I'm loading the following models: await faceapi.loadSsdMobilenetv1Model(); await faceapi.loadFaceLandmarkModel(); await faceapi.loadFaceRecognitionModel();

I tried the following. I take one image with multiple people. I run the face detection. I then save the images of the faces (a little bit unzoomed to see a bit more the person). Then I ran faceapi.detectSingleFace() on each of these images of a single face and it didn't find any face on any of the images. Which is quite interesting since those are actual faces that I found using the same library.

I think the problem comes from the size of the images. The faces were extracted from a picture with many people and the resulting "single face" image are quite small and a bit blurry.

I tried the following 2 things:

• I took one of the single faces images and used gimp to sharpen it. Then the library found a face on it.
• I used an original image with fewer faces on it so each face has a higher definition (https://parismatch.be/app/uploads/2018/10/0000307817-1100x715.jpg) and everything worked fine except for Ross face where I get the same problem, it doesn't find a face from the extracted face picture. I even tried to unzoom a bit to make it bigger but still not finding a face.

I'm a bit surprised that it doesn't find faces at all when it actually extracted from the same library. I guess the models are not the same for both functions.

Thank you very much for all your work, this library is amazing!