Dynamic loss balancing and sequential enhancement for road-safety assessment and traffic scene classification
This work addresses the problem of reducing labor-intensive manual road-safety inspections for transportation authorities, though it appears incremental as it builds on existing neural methods for attribute prediction.
The paper tackles automated road-safety assessment by proposing a two-stage neural architecture that predicts over forty road-safety attributes from video, reducing dependency on manual annotation. The method achieves competitive performance on three datasets, including the iRAP-BH dataset with 2,300 km of labeled road video.
Road-safety inspection is an indispensable instrument for reducing road-accident fatalities contributed to road infrastructure. Recent work formalizes road-safety assessment in terms of carefully selected risk factors that are also known as road-safety attributes. In current practice, these attributes are manually annotated in geo-referenced monocular video for each road segment. We propose to reduce dependency on tedious human labor by automating recognition with a two-stage neural architecture. The first stage predicts more than forty road-safety attributes by observing a local spatio-temporal context. Our design leverages an efficient convolutional pipeline, which benefits from pre-training on semantic segmentation of street scenes. The second stage enhances predictions through sequential integration across a larger temporal window. Our design leverages per-attribute instances of a lightweight bidirectional LSTM architecture. Both stages alleviate extreme class imbalance by incorporating a multi-task variant of recall-based dynamic loss weighting. We perform experiments on the iRAP-BH dataset, which involves fully labeled geo-referenced video along 2,300 km of public roads in Bosnia and Herzegovina. We also validate our approach by comparing it with the related work on two road-scene classification datasets from the literature: Honda Scenes and FM3m. Experimental evaluation confirms the value of our contributions on all three datasets.