Research on Multi-Task Power Large Vision Model Based on Hypernetwork Architecture
Linyu Zhang, Dongmei Cui, Xiaodong Alice Wang, Yiping Zhou, Xiaohang Zhang · 2024
This paper focuses on the research of power large vision model. By carrying out the standardized collection and cleaning of power vision data sets, the feature extraction and representation of multi-task data, and the search and construction of general model architectures for power vision, an integrated process of constructing and training multi-task power large vision model based on the hypernetwork architecture is established. This enables the effective improvement of the inference performance of downstream task models, gives full play to the potential of the model under the existing parameters, and is conducive to the rapid and low-cost implementation of power business scenarios.