Cloud Recognition Target

A cloud recognition target refers to an image that uses cloud-based image recognition.

In Basic AR, Gyroscope AR, Plane AR, and Walk-in AR scenarios, all uploaded target images are considered cloud recognition targets.

What is cloud recognition

Images used as cloud recognition targets should have sharp and detailed features.

For example, a square may contain four feature points, while a circle contains no feature points at all.

When creating a target image, avoid using smooth or soft edges, and avoid images with large areas of solid colors or gradients.

Example of good marker images

Example of bad marker images

If the recognition material is a logo, gesture, color, numbers, or text, it may not be suitable for creating a compliant cloud recognition target image.

In this case, you can click here to contact us and request a custom AI recognition solution.

How to check target image quality

Use star rating as an aid for evaluation

1. After you upload a cloud recognition target image in Basic AR, Gyroscope AR, Plane AR, or Walk-in AR, the system will automatically run a basic check and give a star rating. The star rating is only a rough reference. In general, a higher rating usually means more stable recognition and tracking performance.

2. You can also click here to upload an image and view the target star rating independently.

3. Move your mouse over the target image thumbnail to view its rating. If the rating is below 3 stars, it is recommended to optimize the image based on the cloud recognition target guidelines.

Real-world performance is the final standard

WARNING

The star rating is for reference only. Final performance depends on real-world testing.

In real use, recognition speed and accuracy can be affected by several factors:

• Light around the environment

• Reflections on the image surface

• How clear and different the real object is compared to the uploaded image

• Whether the target image is too big or too small

• Differences in phone cameras and image quality

Extra notes when using the collection feature

1. When using the cloud recognition collection feature, you can open My Collection and check the star rating of each target image on the collection preview page.

2. The target images for each AR scene in the collection should be as different as possible.

If the images are too similar, the system may not be able to determine which AR scene the scanned image belongs to.

For example

The five calendar images have very similar color block layouts, proportions, number shapes, and sizes.

If they are used as the target images for five different scenes in the same collection, it is very likely to cause wrong recognition when using the collection feature.

Guidelines

Type	Description and Illustration
Valide Image Format	JPG/JPEG/PNG/WEBP/BMP/TIF/TIFF/HEIC
Valid Color Mode	Set the image to RGB to help detect any color deviation promptly. When exporting, check Color Space → Convert to sRGB, and verify after export to ensure no color deviation is present.
Recommend Image Resolution	The target image size should be between 480 × 480 px and 1280 × 1280 px. It is recommended to use 800 × 800 px for better recognition results.
Recommend Image Aspect Ratio	For landscape target images, the recommended aspect ratio should be between 1:1 and 16:9. The width divided by height should be in the range of 1 to 1.78. For portrait target images, the recommended aspect ratio should be between 9:16 and 1:1. The width divided by height should be in the range of 0.56 to 1. We strongly recommend using 1:1 square images. They are more balanced and usually have a higher recognition success rate than other formats.
Rich Detail	Cloud recognition target images should contain rich details. Avoid images with large areas of solid color or gradients, or images that are too simple or repetitive.
Avoid extensive whitespace	When creating a target image, it is recommended to crop out large blank areas and low-contrast regions of the image.
Uniform Distribution of Feature Points	Feature points should be evenly distributed across the entire image. Do not let all feature points concentrate in a single area of the image.
High Contrast	Try to increase the contrast of the cloud recognition target image as much as possible.
Avoid Repetitive Pattern	Do not use repetitive or symmetrical patterns as cloud recognition target images. These images may still be recognized, but they can easily cause incorrect recognition or unstable results. In some cases, they may not be recognized at all.
Avoid Image Blur	Avoid designs with large blurred areas.

What if recognition is wrong?

Misrecognition in a single scene

• In Basic AR/Gyroscope, the AR scene appears even when the user is not pointing at the target image.

• In Plane AR and Walk-in AR, scanning is completed even when the user is not pointing at the target image.

Solution

• Improve the uploaded images based on the target image guidelines.

• Basic AR, Gyroscope, Plane AR, and Walk-in AR all allow multiple target images in one scene. Please make sure every uploaded image follows the target image guidelines, and remove any unnecessary or distracting images.

TIP

If misrecognition still occurs, you can improve the result by increasing the Recognition Threshold.

This feature is available for Enterprise users. If you want to enable it, please click here to contact us.

Misrecognition in a collection

When testing a collection, the AR content of Scene B appears even though the target image of Scene A is scanned.

Solution

1. First check Scene B in the same collection. It may also have recognition problems. If Scene B is too easy to trigger, the system may not detect Scene A in time and may match Scene B by mistake.

2. Check if the target images of different scenes in the same collection look too similar. If they look almost the same, the system can get confused and mix them up.

For example, if these five images look very similar, you should add clear differences to each one and remove elements that are too similar. This helps prevent misrecognition from the root.

If misrecognition still happens after adjustments, you can increase the Recognition Threshold to further improve the result.

TIP

If you cannot change the target image due to business needs, you can click here to contact us to enable AI recognition (such as text or logo recognition). This helps the system better distinguish similar images.

What if the image is duplicate?

When you create a new Basic AR/Gyroscope, Plane AR, or Walk-in AR scene and upload a target image, the system will automatically check whether it is very similar to other images in the same collection.

Solution

1. If you do not need the collection multi-image recognition feature, you can click Ignore.

2. If you need to use the collection feature, carefully check whether other scenes in the same collection have similar target images. If you find similar ones, edit or delete them, then upload them again.

Note

If you check and there are no similar images in the collection, you can click Ignore to finish uploading. After uploading, please test it in real use. The final result depends on real testing.