New warning NORM_RECT without IMAGE_DIMENSIONS for face/pose landmarks with 0.10.15

[eyaler]

Have I written custom code (as opposed to using a stock example script provided in MediaPipe)

No

OS Platform and Distribution

windows 64, firefox 130 and chromium 130

MediaPipe Tasks SDK version

0.10.15

Task name (e.g. Image classification, Gesture recognition etc.)

Face landmarks, Pose landmarks

Programming Language and version (e.g. C++, Python, Java)

js

Describe the actual behavior

warning appears regarding NORM_RECT without IMAGE_DIMENSIONS

Describe the expected behaviour

no warning

Standalone code/steps you may have used to try to get what you need

I updated the official face landmarks and pose landmarks examples to 0.10.15:
https://codepen.io/eyaler/pen/abeogLK
https://codepen.io/eyaler/pen/OJKLeOV

Running them gives the warning:
`W0922 18:23:02.646000 1895072 landmark_projection_calculator.cc:189] Using NORM_RECT without IMAGE_DIMENSIONS is only supported for the square ROI. Provide IMAGE_DIMENSIONS or use PROJECTION_MATRIX.`

Need the code in the standard examples be changed?

Other info / Complete Logs

No response

[kuaashish]

Hi @eyaler,

Please allow us some time to review. We will inform you whether it is possible to remove this warning or not.

[kuaashish]

Hi @eyaler,

Could you please verify if the issue is still present in the most recent version 0.10.18 and let us know the status?

Thank you!!

[eyaler]

@kuaashish i updated both codepen examples to 0.10.18. if you run them you will see you get the same warning.

[JasOleander]

Using NORM_RECT without IMAGE_DIMENSIONS is only supported for the square ROI. Provide IMAGE_DIMENSIONS or use PROJECTION_MATRIX. Now I am using 0.10.18 version. but I got the same error.

[arifyaman]

Same here and in python api i think there is no way i can set these without creating a graph :(

[bigger-py]

Hi, I've just updated from mediapipe 0.10.0 to 0.10.21 since 0.10.0 has been pulled from pypi. I'm seeing the same issue on 0.10.21 as you.

I've dug a bit in the source calculator code and subgraphs for pose landmarks and my understanding is that the ROI passed to the landmark projection calculator is always configured to be square.

The ROI comes from either the previous pose-landmarker result (if available) or a new pose-detection result and is forced to be square (see the square_long option last node of the pose landmark to ROI (CPU) and pose detection to ROI (CPU) subgraphs.

I haven't looked at the face landmarker code recently since it's not relevant to my current project but I have in the past and my understanding is that the blazepose detector/landmarker used in Mediapipe was based on the same model architecture used for Mediapipe's face detection, so I would guess the ROI is also square by design for face detection, though you'd need to look an analogous place to confirm that!

Hopefully a mediapipe dev (@kuaashish) can confirm whether the above is valid or I've missed something, but I'm not too worried about the warning for my own work.

[mexicantexan]

I'm also getting this error

[chalapathi-ct]

I am also getting this error.

[one-matrix]

I'm also getting this error

[VeeranjaneyuluToka]

I am also getting the same error.

[andypotato]

Getting the same error on 0.10.21.

[MikhailIvanou]

+1

[bpguillem]

+1

[Picculi]

+1

[qiufuyu123]

+1

[MarvinRucinski]

I just checked multiple versions od media pipe this error does not appear in mediapipe 0.10.14. All the versions above 0.10.14 have this error. Also in hands landmarker.

[jaffe-fly]

+1

[jbartolozzi]

+1 version 0.10.21

[japizarroc]

Same error in Successfully installed mediapipe-0.10.21

[japizarroc]

error INFO: Created TensorFlow Lite XNNPACK delegate for CPU. WARNING: All log messages before absl::InitializeLog() is called are written to STDERR W0000 00:00:1751762859.332325 11556 inference_feedback_manager.cc:114] Feedback manager requires a model with a single signature inference. Disabling support for feedback tensors. W0000 00:00:1751762859.362026 11556 inference_feedback_manager.cc:114] Feedback manager requires a model with a single signature inference. Disabling support for feedback tensors. W0000 00:00:1751762864.019622 11556 landmark_projection_calculator.cc:186] Using NORM_RECT without IMAGE_DIMENSIONS is only supported for the square ROI. Provide IMAGE_DIMENSIONS or use PROJECTION_MATRIX.

[japizarroc]

ERROR UPDATE... now it pops only W0000 00:00:1751863009.707015 8276 landmark_projection_calculator.cc:186] Using NORM_RECT without IMAGE_DIMENSIONS is only supported for the square ROI. Provide IMAGE_DIMENSIONS or use PROJECTION_MATRIX.

[EugeniaLevy]

ERROR UPDATE... now it pops only W0000 00:00:1751863009.707015 8276 landmark_projection_calculator.cc:186] Using NORM_RECT without IMAGE_DIMENSIONS is only supported for the square ROI. Provide IMAGE_DIMENSIONS or use PROJECTION_MATRIX.

How did you get rid of the ones before?

[japizarroc]

I just changed Python version to 3.12.1