Training method of work content recognition model, work content recognition method and device
By defining task description data and sample data, generating sample labels using a label generation model, and conducting cross-modal training, a target task content recognition model is constructed. This solves the problem of unified processing of task content in different modalities and achieves efficient recognition and management.
CN122153454APending Publication Date: 2026-06-05BEIJING YUANLI WEILAI SCI & TECH CO LTD
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- BEIJING YUANLI WEILAI SCI & TECH CO LTD
- Filing Date
- 2026-03-18
- Publication Date
- 2026-06-05
Smart Images

Figure CN122153454A_ABST
Abstract
Embodiments of the present specification provide a job content recognition model training method, a job content recognition method and device, wherein the job content recognition model training method comprises: determining task description data and sample data of a job content recognition task, wherein the sample data comprises image sample data and text sample data; inputting the sample data and the task description data into a label generation model to obtain sample labels corresponding to the sample data output by the label generation model, wherein the sample labels comprise first sample labels corresponding to the image sample data and second sample labels corresponding to the text sample data; determining target training data according to the sample data and the sample labels, and performing cross-modal training on an initial job content recognition model using the target training data to obtain a target job content recognition model, wherein the target job content recognition model is used to perform the job content recognition task. The recognition ability of the model for different modal job content is improved.
Need to check novelty before this filing date? Find Prior Art