How to ignore head repairs that are less than 5 percent of the total image?

[joseph16388]

How do I ignore head fixes that are less than 5 percent of the total image through the API? thanks！

[dosubot[bot]]

To ignore head repairs that are less than 5 percent of the total image through the API, you can use the ad_mask_min_ratio slider to set the minimum mask area ratio. Setting this value to 0.05 will ensure that any repairs less than 5 percent of the total image are ignored.

Here is the relevant code snippet:

w.ad_mask_min_ratio = gr.Slider(
    label="Mask min area ratio" + suffix(n),
    minimum=0.0,
    maximum=1.0,
    step=0.001,
    value=0.05,  # Set this value to 0.05 to ignore repairs less than 5 percent
    visible=True,
    elem_id=eid("ad_mask_min_ratio"),
)

This configuration will ensure that only repairs that cover at least 5 percent of the total image area are considered ^[1][2].

To continue talking to Dosu, mention @dosu.

[joseph16388]

What is the default value of ad_mask_min_ratio in the adetailer api? @dosu

[dosubot[bot]]

The default value of ad_mask_min_ratio in the adetailer API is 0.0. This means that, by default, there is no minimum mask area ratio, and all repairs are considered regardless of their size.

To continue talking to Dosu, mention @dosu.

[joseph16388]

The default value of ad_mask_min_ratio in the adetailer API is 0.0. This means that, by default, there is no minimum mask area ratio, and all repairs are considered regardless of their size.

To continue talking to Dosu, mention @dosu.

However, when I did not set it, I found that some relatively small faces were ignored. Is it because this relatively small face was not detected during face detection, so is there a threshold for face size detection? @dosu

[dosubot[bot]]

Yes, it is possible that small faces were not detected during face detection due to a threshold for face size detection in the adetailer API. The mediapipe_face_detection function uses the min_detection_confidence parameter, which is set to a default value of 0.3. This parameter can influence the detection of smaller faces, as lower confidence detections might be filtered out.

with mp_face_detection.FaceDetection(
    model_selection=model_type, min_detection_confidence=confidence
) as face_detector:
    pred = face_detector.process(img_array)

If the confidence score of the detected faces is below this threshold, they will not be included in the output, potentially leading to smaller faces being missed ^[1].

To continue talking to Dosu, mention @dosu.