- jpg
- jpeg
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.
Landscape: 1:1 to 16:9 (Aspect Ratio 1 to 1.78)
Portrait: 9:16 to 1:1 (Aspect Ratio 0.56 to 1)
- The following are examples of poor-quality images:
- For the cases above, you may use AI Recognition instead of Cloud Recognition.
Minimize blank regions and ensure the main subject is clearly emphasized.
- The image should contain sharp, well-defined details:
A square typically contains around 4 feature points, while a circle contains almost none.
Therefore, avoid large circular shapes in the design whenever possible.
- Avoid soft, smooth edges and images with large gradient areas.
Increasing the image contrast can significantly improve recognition accuracy.
Original:
Contrast Enhanced:
Recognition:
- More contrast examples:
- The following repetitive images can technically be recognized, but they carry a higher risk of misrecognition
(If each repeated object contains rich detail, recognition may still be possible):
- The following images, which are repetitive, symmetrical, and overly simple, cannot be recognized
(Symmetry reduces recognizability, and the lines are too uniform):
Avoid large areas of blur in the design whenever possible
Original:
Large Area of Blur:
Recognition:
- The image should have a uniform distribution of features.
- As shown above, when only the lower-right corner contains dense details, it should be avoided.
The following example shows feature points concentrated in the lower-left corner - this should also be avoided: