Fast Positioning Visual Abnormalities after Inputting Images of Different Sizes

Symptom:

After inputting images of different sizes, visual abnormalities occurred in the fast positioning module, leading to positioning errors.

As shown below, the left figure shows the normal visualized result, while the right one shows the abnormal visualized result.

Affected Software Versions:

Mech-DLK 2.4.2 and earlier versions

Possible Causes:

The size of the input image during deep learning model package inference was different from the size during model training, which led to positioning errors.

Solution:

Use Mech-DLK 2.5.0 to export new fast positioning model package, and then conduct the model package inference.