移动端实时交互式分割方法、装置、存储介质及电子设备
By combining a backbone network and a gated convolutional structure with a separable convolutional decoder, and employing a hybrid training method of click and smear, the problem of interactive segmentation where users can independently select any object on mobile devices is solved, achieving convenient and efficient image segmentation results.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HANGZHOU QUWEI SCI & TECH
- Filing Date
- 2023-01-06
- Publication Date
- 2026-07-17
AI Technical Summary
Existing intelligent image segmentation algorithms struggle to enable interactive segmentation of any object selected by the user on mobile devices, and the click or selection method is inconvenient to operate on small interactive interfaces.
Employing a backbone network and gated convolutional structure, combined with a separable convolutional decoder, and using a hybrid training method of click and smudge, it achieves real-time interactive segmentation on mobile devices. Utilizing a semantic segmentation dataset and randomized cyclic training, it supports both click and smudge interaction methods.
A user-friendly, real-time interactive segmentation feature was implemented on a small interactive interface, improving the convenience of image editing and the accuracy of segmentation results.
Smart Images
Figure CN116129116B_ABST