Indoor Semantic Segmentation Dataset

Published: 14 September 2021| Version 1 | DOI: 10.17632/hs5w7xfzdk.1
Elhassan Mohamed


Folders descriptions: - images -> dataset images. - PixelLabelData -> pixels representation for each image (annotations). - -> a compressed folder of the dataset files and folders. Files descriptions: - imds.mat -> images datastore in Matlab format. - pxds.mat -> pixels datastore in Matlab format. Note: files paths in the imds and pxds Matlab files need to be modified to point to the new location of the images on your disk. Classes: - Classes -> 'door' 'pull door handle' 'push button' 'moveable door handle' 'push door handle' 'fire extinguisher' 'key slot' 'carpet floor' 'background wall' Video: - IndoorDataset_960_540.mp4 -> a short video that can be used to benchmark the system's speed (FPS) and inference capabilities.


Steps to reproduce

The dataset is collected by driving a powered wheelchair through the indoor environment corridors. A one-minute video is recorded and annotated. To import the dataset for training into Matlab, load the image data store 'imds' and the pixel label datastore 'pxds' into Matlab. Please remember to change the folders’ path in imds and pxds. Raw images and raw annotated pixels can be used with any other platform. Objects of interest are mentioned in the 'readMe' file.