The Semantic Segmentation or Pixel-level labeling is used to label each and every pixel in the image. Unlike polygonal segmentation devised specifically to detect a defined object of interest, full semantic segmentation provides a complete understanding of every pixel of the scene in the image. This kind of for detection and localization of specific objects in pixel-level.
There are three main use cases in semantic segmentation:
- Instance segmentation: Used for training perception models in non-environmental objects of interest
- Full pixel semantic segmentation: Used in autonomous vehicles and safety surveillance cameras where information of every pixel is critical and may influence the accuracy of the perception model.
- Panoptic segmentation: Used to individually segment objects of the same class by assigning instance unique IDs to each object.
Complex cases handled by Playment:
- Dense Vegetation (with scattered patches of the sky) and differentiating between terrain and vegetation
- Occlusion related cases – Vegetation occluding building/ vehicle/ pedestrian
- Wires across the sky
- Lanes and signs
As semantic segmentation involves pixel-wise accuracy, it has the highest time and cost per annotation. Playment is a fully managed data labeling platform generates high-quality annotations at scale. We combine both machine learning and human intelligence in a single platform for enabling better autonomous vehicle systems.
Know more about the Image annotation services by Playment here.