CueCAn: Cue Driven Contextual Attention Unit for Identifying Missing Traffic Signs on Uncontrained Roads

Inference in Colab

https://colab.research.google.com/drive/1ujnqXKmQHUQP91COmuu_HXkjitzh0fGC?usp=sharing

Please Note: The accompanied video has been compressed using the HEVC Codec, and may fail to load on Windows media player, and MacOS's default player. Please run the video on VLC player, which comes with the codec support by default. Alternately, the video can also be found here

Conda Environment Containing All Required Dependecies

''' conda env create --name fcnkeras --file=environment.yml '''

Missing Traffic Signs Video Dataset Samples

Missing Signs

Missing Traffic Signs Video Dataset (MTSVD)

Marked Missing Intervals

Traffic Sign Annotations and Tracks

CueCAn GradCam

Video Result