17807614. PANOPTIC SEGMENTATION WITH PANOPTIC, INSTANCE, AND SEMANTIC RELATIONS simplified abstract (QUALCOMM Incorporated)
PANOPTIC SEGMENTATION WITH PANOPTIC, INSTANCE, AND SEMANTIC RELATIONS
Organization Name
Inventor(s)
Shubhankar Mangesh Borse of San Diego CA (US)
Hyojin Park of San Diego CA (US)
Debasmit Das of San Diego CA (US)
Risheek Garrepalli of San Diego CA (US)
Fatih Murat Porikli of San Diego CA (US)
PANOPTIC SEGMENTATION WITH PANOPTIC, INSTANCE, AND SEMANTIC RELATIONS - A simplified explanation of the abstract
This abstract first appeared for US patent application 17807614 titled 'PANOPTIC SEGMENTATION WITH PANOPTIC, INSTANCE, AND SEMANTIC RELATIONS
Simplified Explanation
The abstract describes a new framework for combining semantic and instance contexts for panoptic segmentation in image processing. The method involves generating a panoptic encoding by processing semantic and instance feature data using a panoptic encoding generator. The panoptic encoding is then processed to generate panoptic segmentation features, which are used to generate a panoptic segmentation mask.
- The framework integrates semantic and instance contexts for panoptic segmentation.
- It involves processing semantic and instance feature data to generate a panoptic encoding.
- The panoptic encoding is then used to generate panoptic segmentation features.
- These features are used to generate a panoptic segmentation mask.
Potential Applications
This technology has potential applications in various fields, including:
- Computer vision
- Image processing
- Object recognition
- Autonomous vehicles
- Robotics
Problems Solved
The framework addresses the following problems:
- Difficulty in accurately segmenting objects in images
- Challenges in combining semantic and instance contexts for segmentation
- Lack of efficient methods for generating panoptic segmentation masks
Benefits
The benefits of this technology include:
- Improved accuracy in segmenting objects in images
- Enhanced integration of semantic and instance contexts for segmentation
- Efficient generation of panoptic segmentation masks
Original Abstract Submitted
Aspects of the present disclosure relate to a novel framework for integrating both semantic and instance contexts for panoptic segmentation. In one example aspect, a method for processing image data includes: processing semantic feature data and instance feature data with a panoptic encoding generator to generate a panoptic encoding; processing the panoptic encoding to generate a panoptic segmentation features; and generating the panoptic segmentation mask based on the panoptic segmentation features.