Size mismatch between label and color lists in CObjectDetectionTask
LudoBar opened this issue · comments
Description
Using object detection algorithm like GroundingDino leads to memory corruption in vectors storing labels and colors.
Indeed, GroundingDino outputs dynamic list of labels depending on the source image. In some scenario, vectors for labels and colors doesn't have the same size and leads to an index out of bound error (not handled).
Expected behavior
These two data structure must always have the same size, design must guarantee this and implement valid error handling if it is not the case.