Object Detection Is not Working Well With Custom TFLite model from MediaPipe Model Maker

[darrenwong12]

Describe your issue. If applicable, add screenshots to help explain your problem.

I am currently implementing object detection and pose detection in the same camera frame using image stream. I am using a custom tflite model for object detection with input size as 384x384. The realtime performance isn’t so great was just wondering if it’s because for camera stream the resolution is set to high with a resolution of 1280x 720. How is the image resize to passed into my detection model. Is it resized with padding ?

Steps to reproduce.

NIL

What is the expected result?

Better Inference on bounding boxes

Did you try our example app?

Yes

Is it reproducible in the example app?

Yes

Reproducible in which OS?

iOS and Android

Flutter/Dart Version?

[√] Flutter (Channel stable, 3.27.3, on Microsoft Windows [Version 10.0.22631.4890], locale en-SG)
[√] Windows Version (Installed version of Windows is version 10 or higher)
[√] Android toolchain - develop for Android devices (Android SDK version 36.0.0-rc4)
[√] Chrome - develop for the web
[√] Visual Studio - develop Windows apps (Visual Studio Community 2022 17.12.4)
[!] Android Studio (version 2021.3)
X Unable to find bundled Java version.
[√] Android Studio (version 2024.2)
[√] VS Code (version 1.97.2)
[√] Connected device (3 available)
[√] Network resources

Plugin Version?

google_ml_kit: ^0.19.0
camera_android: ^0.10.9+17
camera: ^0.11.0+2
image_picker: ^1.1.2
path_provider: ^2.1.4
google_mlkit_commons: ^0.9.0
google_mlkit_pose_detection: ^0.13.0