INSTANCE SEGMENTATION
In contrast to semantic segmentation, where each pixel is assigned a class, instance segmentation also considers the individual instance of each object. This means that even multiple instances of the same object class (e.g., multiple apples) are identified and segmented precisely and independently.
A deep learning model is trained with a sufficient amount of training data to recognize and segment these precise object boundaries. The bounding box for localization is combined with pixel-perfect segmentations to ensure a precise separation between objects.
INSTANCE SEGMENTATION
High Precision: Through pixel-perfect segmentation, even overlapping or adjacent objects are detected accurately.
Mastering Complex Scenarios: Particularly effective for object detection of multiple objects in an image that touch or overlap.
Easy Integration: Seamless integration into HALCON and MERLIC for industrial applications with minimal programming effort.
Flexibility and Adaptability: Quick adaptation to new scenarios without the need for extensive model retraining.
INSTANCE SEGMENTATION
In the medical and pharmaceutical industries, instance segmentation is used for the precise measurement of bacteria in Petri dishes or particles in liquids. The technology helps determine the exact number and position of cells or structures.
In bin picking, where randomly arranged objects are grasped from boxes, instance segmentation enables precise detection of objects that overlap or are closely packed together. This technology is key to automation in logistics and manufacturing.
