Cloud Recognition Image

Type

Description and Illustration

Valide Image Format

• jpg

• jpeg

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 image resolution is recommended to be between 480×480 and 1280×1280, with around 800 being ideal.

Recommend Image Aspect Ratio

Landscape: 1:1 to 16:9 (Aspect Ratio 1 to 1.78)

Portrait: 9:16 to 1:1 (Aspect Ratio 0.56 to 1)

Rich Detail

• The following are examples of poor-quality images:

• For the cases above, you may use AI Recognition instead of Cloud Recognition.

Avoid extensive whitespace

Minimize blank regions and ensure the main subject is clearly emphasized.

Basic Principles of Cloud Recognition

• 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.

Enhance Contrast

Increasing the image contrast can significantly improve recognition accuracy. Original:

Contrast Enhanced:

• More contrast examples:

Avoid Repetitive Pattern

• 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 Image Blur

Avoid large areas of blur in the design whenever possible

Original:

Large Area of Blur:

Recognition:

Uniform Distribution of Feature Points

• 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: