Research on automatic annotation and classification of video content based on machine learning
The Convolutional Mini-batch Gradient (CMG) architecture represents a pioneering synthesis of Convolutional Neural Networks (CNNs) and the Mini-batch Gradient Descent (MGD) training technique. This amalgamation leverages the inherent spatial feature extraction prowess of CNNs and the computational e...
| 要約: | The Convolutional Mini-batch Gradient (CMG) architecture represents a pioneering synthesis of Convolutional Neural Networks (CNNs) and the Mini-batch Gradient Descent (MGD) training technique. This amalgamation leverages the inherent spatial feature extraction prowess of CNNs and the computational efficiency of MGD, ushering in a novel approach to the automated annotation and categorization of video content. At the heart of CMG lies the utilization of CNNs to conduct a thorough excavation of visual features embedded within individual video frames. Concurrently, the Mini-batch Gradient Descent strategy accelerates the iteration of the model, facilitating faster learning cycles and enhancing the model's adaptability. This dual-pronged approach yields a significant boost in the precision and recall rates associated with video analysis, thereby enriching the comprehensiveness and accuracy of the resultant annotations. Through a series of empirical evaluations and comparisons, the CMG framework has not only demonstrated a considerable augmentation in the resilience and computational efficacy of the model when operating in intricate and variable environments but has also substantiated its reliability and practicality in the realm of automated video content labeling and classification. The model's ability to maintain high levels of performance despite environmental complexities is a testament to its robust design. Furthermore, CMG's proven track record in enhancing computational efficiency and delivering precise outcomes positions it as a viable and promising solution for advancing the frontiers of video content understanding and analysis in the era of big data and machine learning. © 2024 IEEE. |
|---|---|
| ISSN: | |
| DOI: | 10.1109/ICCASIT62299.2024.10827856 |