Question:

[orangecoding]

Hi,

Thanks for this awesome lib!

I'm using it to extract faces from photos, check if they are tilted, if they are, rotate the canvas to straighten the face and then store it as a new image.

I'm basically doing it like this:

    const c = await image(inputPath);
    const face = await detectFace(c);
    const locations = await faceapi.locateFaces(c);
    const faceImg = await faceapi.extractFaces(c, locations);
    const final = canvas.createCanvas(face.landmarks.imageWidth, face.landmarks.imageHeight);

    const roll = face.angle.roll;
    final.getContext('2d').rotate(roll * Math.PI / 180);
    final.getContext('2d').drawImage(faceImg[0], 0, 0);
    writeFileToDirectory(outputPath, final.toBuffer('image/jpeg'));

I'm having 2 questions.

• Do I rly have to call faceapi.locateFaces(c) to get the locations to extract the face?
• Is it somehow possible, to make all faces the same size?

My endgoal is to have all faces the same size (e.g 800x600) later on

[vladmandic]

locateFaces is a reduced version of detectAllFaces, so in your cases just use locateFaces, you shouldn't need anything before
for resizing, there are tons of options, but since you already have a canvas you've created, you can simply resize it fore writing it to file

[vladmandic]

closing due to inactivity, but can be reopened if additional information is needed

[orangecoding]

Thanks :)