Emergency Escape Ramp Scene Dataset

Description

Our dataset comprises 500 RGB-D images, where the depth maps were generated using the AdelaiDepth (LeReS) framework. AdelaiDepth employs a relative depth learning strategy, focusing on encoding the ordinal relationships between pixels. It also applies scale and shift normalization to align the depth values, thereby mitigating depth-scale ambiguity. All images are resized to 512×512 pixels. The dataset contains nine classes: background, emergency escape ramp, escape ramp sign, road surface, emergency lane, guardrail, no parking sign, and person. Out of the total dataset, 450 images are allocated for training, while the remaining 50 are reserved for testing.

Institutions

Categories

Licence