FaceQualityCNN

Face Quality CNN

This model is capable of evaluating the quality of human faces in input images.

Inference

It takes aligned face images (112x112) as input and returns a quality score for the face (ranging from 0 to 1, with higher scores indicating higher quality).

Code

class face_quality_assessment():
    def __init__(self, path):
        # Initialize model
        self.net = cv2.dnn.readNet(path)
        self.input_height = 112
        self.input_width = 112

    def classify(self, srcimg):
        input_img = cv2.resize(cv2.cvtColor(srcimg, cv2.COLOR_BGR2RGB), (self.input_width, self.input_height))
        input_img = (input_img.astype(np.float32) / 255.0 - 0.5) / 0.5

        blob = cv2.dnn.blobFromImage(input_img.astype(np.float32))
        self.net.setInput(blob)
        outputs = self.net.forward(self.net.getUnconnectedOutLayersNames())
        return outputs[0].reshape(-1)

fqa = face_quality_assessment("weights/face-quality-assessment.onnx")
fqa_probs = fqa.classify(img)

---

license: apache-2.0