Abnormality detection and usability verification method, device, medium and product of body data
By aligning multimodal embodied data with timestamps, video quality, data integrity, temporal synchronization, and the rationality of state actions are detected. This solves the problem of the lack of automated quality detection in existing technologies, and enables efficient and accurate anomaly identification and cleaning of embodied data, ensuring the stability and learning effect of the embodied intelligent model.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHANGHAI COOPERS TECHNOLOGY CO LTD
- Filing Date
- 2026-04-28
- Publication Date
- 2026-06-19
AI Technical Summary
Existing technologies lack unified, systematic, and automated quality inspection methods for multimodal embodied data, making it difficult to comprehensively identify hidden anomalies in multiple dimensions such as video, time series, and state actions. This results in defective embodied data directly flowing into the model training process, affecting the learning effect and system stability of embodied intelligent models.
By using timestamps as a unified time reference, multimodal perception data, state data, and action data are time-aligned, temporal correspondences are extracted, and anomaly detection processing is performed, including video quality detection, data integrity detection, temporal synchronization detection, and state/action rationality detection. Finally, the detection results are fused to generate a comprehensive evaluation result.
It achieves accurate identification and automated cleaning of multi-dimensional anomalies, ensuring the reliability of the data entering the database, guaranteeing the high quality and security of the training data for embodied intelligent models, and reducing the time and cost of manually checking for abnormal data.
Smart Images
Figure CN122241155A_ABST