The Matlab codes in this repository are the implementation of automatic labeling approach for asynchronous events data proposed in paper "A Neuromorphic Dataset for Object Segmentation in Indoor Cluttered Environment".

Cite the article:

Huang, X., Sanket, K., Ayyad, A., Naeini, F. B., Makris, D., & Zweiri, Y. (2023). A Neuromorphic Dataset for Object Segmentation in Indoor Cluttered Environment. arXiv preprint arXiv:2302.06301.

Dataset repository:

Download dataset Here

Visualization

RGB frame:

Annotated mask:

Annotated events:

Dataset structure:

|-- ESD-1 (training)
    |-- conditions_1
        |-- RGB
            |-- images                   (raw RGB images)
            |-- masks                    (annotated masks)
        |-- events
            |-- left.mat                 (events info from left event camera, RGBD info, camera's movement info)
            |-- right.mat                (events info from right event camera, RGBD info, camera's movement info)
            |-- events_frame.mat         (synchronous left and right image frames tranformed from RGBD coordinate)
            |-- mask_events_frame.mat    (synchronous left and right masks tranformed from RGBD coordinate)
    |-- conditions_2
        |-- ...
    ...

|-- ESD-2 (testing)
    |-- conditions_1
        |-- RGB
            |-- images
            |-- masks
        |-- events
            |-- left.mat
            |-- right.mat
            |-- events_frame.mat
            |-- mask_events_frame.mat
    |-- conditions_2
        |-- ...
    ...

Getting Started

execute the labeling_auto_clean_2.m file in Matlab

events information is stored in "event_labeled" variable in left.mat and right.mat