- Media recognition
Instance Segmentation
We developed a technology that detects and declinates various object in an image.
- Simultaneously estimates fields separated by object type and regions representing individual objects, thereby saving memory while at the same time increasing accuracy.
- Used as a shape estimation function to enable robots to identify objects.
Applications
- Unloading robots and picking robots on automated logistics lines
- Sensing of the surrounding environment for automatic driving
Benchmarks, strengths, and track record
- Public benchmark data for instance segmentation; Achieved the world’s best accuracy using Pascal VOC.
As of September 2017; Source: Toshiba
Inquiries
Please include the title “Toshiba AI Technology Catalog: Object field extraction” or the URL in the inquiry text.
Please note that because this technology is currently the subject of R&D activities, immediate responses to inquiries may not be possible.
References:
- Viet Pham, et al., “BiSeg: Simultaneous Instance Segmentation and Semantic Segmentation with Fully Convolutional Networks,” 28th British Machine Vision Conference, 2017.
- Toshiba Clip: Toshiba’s robot solutions for a new era (Pt. 1: Technology) (in Japanese)
- Developed AI that estimates individual cargo fields from regular camera images with the world’s highest level of accuracy
- S. Ito and S. Kubota, “Point Proposal based Instance Segmentation with Rectangular Masks for Robot Picking Task”, ACCV2020.