About this document
Attention-Guided Risky Object Localization by seba.mohamed1 is a document available to read on EtoBox.
This paper presents an Attention-guided Multistream Feature Fusion Network (AM-Net) designed to localize risky traffic agents in driving videos captured by dashcams. The proposed method utilizes Gated Recurrent Units (GRUs) to analyze spatio-temporal features and introduces a new dataset, Risky Object Localization (ROL), to support its development. AM-Net achieves a state-of-the-art performance of 85.73% AUC on the ROL dataset, demonstrating its effectiveness in predicting and localizing potential traffic r
- Author
- seba.mohamed1
- Language
- EN