landmarks-regression-retail-0009: Question on the output for this model

[anuragrawal]

For the following model : https://docs.openvino.ai/2024/omz_models_model_landmarks_regression_retail_0009.html, there's a statement for outputs:  All the coordinates are normalized to be in range [0, 1]

How does this normalization happen? - By dividing the x-coordinates by width of the image and y-coordinates by height of the image? I am referring to face recognition demo but am using another landmarks model. That model doesn't return normalized coordinates so I would need to mimic your normalization logic so that the face identification/recognition part can work properly.

Let me know in case of any questions. Thanks!

[Hairul_Intel]

Hi anuragrawal,

Thank you for reaching out to us.

We're checking this with the relevant teams and will update you as soon as possible.

Regards,

Hairul

[anuragrawal]

@Hairul_IntelDo you have an update for me? Thanks!

[Hari_B_Intel]

Hi anuragrawal

Sorry for the late response, to answer your question, there is some useful information from https://docs.openvino.ai/2024/omz_models_model_face_reidentification_retail_0095.html

To normalize coordinates, we divide coordinates in pixels for the x- and y-axis by the image's width and height.

The face image will aligned in five-point sequences of the left eye, right eye, tip of nose, left lip corner, and right lip corner, which are located in the following points in normalized coordinates [0,1]x[0,1]:

[(0.31556875000000000, 0.4615741071428571),

(0.68262291666666670, 0.4615741071428571),

(0.50026249999999990, 0.6405053571428571),

(0.34947187500000004, 0.8246919642857142),

(0.65343645833333330, 0.8246919642857142)]

[Peh_Intel]

Hi anuragrawal,

This thread will no longer be monitored since we have provided an answer. If you need any additional information from Intel, please submit a new question. 

Regards,

Peh