Search Issue Tracker
Won't Fix
Votes
0
Found in [Package]
5.1.0-pre.10
5.1.0
6.0.0-pre.3
Issue ID
ARFB-360
Regression
No
[Android] Portrait Images are rotated by 90 degrees when tracked on Android
Reproduction steps:
1. Open the attached project “ARProject”
2. Open the “Assets/Scenes/ImageTracking/Images/Rafflesia.jpg” Image
3. Build And Run on Android
4. In the app, select the "Tracking/Basic Image Tracking" button
5. Track the "Rafflesia.jpg" image with the Android device
Expected result: The tracked image rotation matches the real image rotation exactly
Actual result: The tracked Image is rotated by 90 degree
Reproducible with: 5.1.0-pre.10, 5.1.0(2021.3.31f1, 2022.3.9f1, 2023.1.13f1, 2023.2.0b9, 2023.3.0a6), 6.0.0-pre.3(2023.2.0b9, 2023.3.0a6)
Couldn't test with 4.2.8, 5.0.7 (2021.3.31f1, 2022.3.9f1, 2023.1.13f1, 2023.2.0b9, 2023.3.0a6) (AR foundation script errors)
Reproducible with these devices:
VLNQA00518 - Google Pixel 4 (Pixel 4), CPU: Snapdragon 855 SM8150, GPU: Adreno 640, OS: 12
VLNQA00414 - Galaxy Note10+ 5G (SM-N976V), CPU: Snapdragon 855 SM8150, GPU: Adreno 640, OS: 9
VLNQA00278 - Xiaomi Redmi Note 7 (Redmi Note 7), CPU: Snapdragon 660, GPU: Adreno 512, OS: 9.0.0
Testing environment: macOS(Intel) 14.0
Add comment
All about bugs
View bugs we have successfully reproduced, and vote for the bugs you want to see fixed most urgently.
Latest issues
- Rendering locks up when not looking at a transparent material on Meta Quest 2
- Volumetrics break when using a Custom Pass to create a Thickness Buffer for Alpha Clipping
- All tests are run instead of only the failed ones when the "Rerun Failed" button is pressed
- GameObject is not masked when the "Render PostProcessing Effects" pass executes with a resolved non-MSAA Color target and MSAA DepthStencil target
- [Android] Gfx.WaitForGfxCommandsFromMainThread high performance usage and inconsistency when built Player scene has Canvas GameObject on a specific Project
Resolution Note:
After testing different portrait images, we observed that images with higher quality scores are tracked properly in portrait orientation. Images with very low quality scores often have problems being tracked properly by ARCore and/or ARKit. This can cause the underlying AR platforms to mixup the image pose or dimensions that are communicated to ARFoundation. It is recommended to use ARCore’s arcoreimg tool to evaluate your images in their target orientations and make sure they have a sufficiently high quality score (at least 75 is recommended by ARCore).
https://developers.google.com/ar/develop/augmented-images/arcoreimg