[Code] Design of resilient smart highway systems with data-driven monitoring from networked cameras
Description
Traditional high-way transportation systems are monitored based on traffic counters. Such sensors provide much less information compared to traffic cameras and make the system less secure/resilient to attacks/disasters. Thanks to the success of deep learning for object detection/segmentation on images and the publicly available large-scale image datasets with object labels, fusing the information from both traffic counters and traffic cameras has the potential to improve the security and resilience of existing high- way transportation systems. The purpose of the project is to investigate such a potential by developing a deep-learning-based highway video monitoring method that can reliably estimate the fine-grained (car/truck/motorcycle) traffic flow of a high-way network. First, we need to collect a large- scale traffic video dataset with traffic flow estimations from corresponding traffic counters. Then, we need to find efficient deep learning methods for extracting fine-grained local traffic information from individual traffic videos. At last, we need to correlate this information with traffic counters for sensor fusion and detection of defective counters.
Files
CornerNet-Lite.zip
Files
(2.9 GB)
Name | Size | Download all |
---|---|---|
md5:989dd4326df32ed67bb43f653298017a
|
2.3 GB | Preview Download |
md5:796ed667266cfedb9562168bb6f445f7
|
532.3 MB | Preview Download |