首页|Rochester Institute of Technology Researcher Updates Current Study Findings on M achine Learning (Regulating Modality Utilization within Multimodal Fusion Networ ks)
Rochester Institute of Technology Researcher Updates Current Study Findings on M achine Learning (Regulating Modality Utilization within Multimodal Fusion Networ ks)
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NETL
NSTL
A new study on artificial intelligence is now available. According to news originating from Rochester, New York, by Ne wsRx correspondents, research stated, "Multimodal fusion networks play a pivotal role in leveraging diverse sources of information for enhanced machine learning applications in aerial imagery." Financial supporters for this research include National Geospatial-intelligence Agency; Air Force Office of Scientific Research; National Science Foundation. Our news reporters obtained a quote from the research from Rochester Institute o f Technology: "However, current approaches often suffer from a bias towards cert ain modalities, diminishing the potential benefits of multimodal data. This pape r addresses this issue by proposing a novel modality utilizationbased training method for multimodal fusion networks. The method aims to guide the network's ut ilization on its input modalities, ensuring a balanced integration of complement ary information streams, effectively mitigating the overutilization of dominant modalities. The method is validated on multimodal aerial imagery classification and image segmentation tasks, effectively maintaining modality utilization withi n ±10% of the user-defined target utilization and demonstrating th e versatility and efficacy of the proposed method across various applications. F urthermore, the study explores the robustness of the fusion networks against noi se in input modalities, a crucial aspect in real-world scenarios. The method sho wcases better noise robustness by maintaining performance amidst environmental c hanges affecting different aerial imagery sensing modalities."
Rochester Institute of TechnologyRoche sterNew YorkUnited StatesNorth and Central AmericaCyborgsEmerging Tech nologiesMachine Learning
NETL
NSTL
