Real-time online action detection method based on mamba architecture

By adopting an online action detection method based on the Mamba architecture, which combines an action detection module and a future prediction module, the problems of insufficient long-term dependency modeling and real-time computing resource limitations are solved, achieving efficient and accurate action detection, especially action boundary recognition in complex scenarios.

CN120412076BActive Publication Date: 2026-04-17SOUTHEAST UNIV +1
0 Cites 0 Cited by

Patent Information

Application Number
CN202510426450.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2025-04-07
Publication Date
2026-04-17
Estimated Expiration
2045-04-07

AI Technical Summary

Technical Problem

Existing online action detection methods are insufficient in effectively modeling long-term dependencies, making it difficult to accurately capture action features and dependencies over long periods in videos. Real-time inference is challenging under limited computing resources, and their ability to recognize ambiguous action boundaries in complex scenes is also insufficient.

Method used

An online action detection method based on the Mamba architecture is adopted. By combining the action detection module (ADM) and the future prediction module (FAM), the method captures spatiotemporal dependencies using selective state space and convolution mechanisms. Combined with linear projection layers and an improved Mamba block design, efficient action detection is achieved.

Benefits of technology

It improves the accuracy and robustness of action detection, reduces false positives, meets the needs of real-time applications, enhances the ability to identify action boundaries in complex scenarios, and has higher detection accuracy and generalization ability.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The application provides a real-time online action detection technology based on a Mamba architecture, comprising: S1, training data acquisition, acquiring video streams of multiple perspectives in a target scene; S2, data preprocessing, extracting video features from the video streams based on a pre-trained two-dimensional convolutional neural network; S3, introducing a Mamba OAD framework, which is composed of an action detection module and a future prediction module; S4, constructing the action detection module, which is mapped to a feature space through a linear projection layer, and the linear projection layer converts input features to D dimensions through a learnable weight matrix; S5, constructing the future prediction module, which predicts future latent features based on short-term latent features output by the action detection module; and S6, model training and reasoning, in which related modules are jointly optimized in the training process. The mAP and mcAP indicators are used to measure the online action detection accuracy, and the FPS and GFLOPs indicators are used to measure the complexity of model inference. The application improves the accuracy and robustness of action detection.
Need to check novelty before this filing date? Find Prior Art