DoorDetect is a dataset of 1,213 images that have been annotated with object bounding boxes. The images are very diverse and often contain complex scenes with several objects.
The images annotated are from Open Images Dataset V4 and MCIndoor20000 .
The identified object classes are: handle; door, which refers to any room door; cabinet door, which refers to any drawer or small door; and refrigerator door, which refers to any door in a refrigerator.
The object location is specified by the coordinates of its bounding box. Boxes were marked using Yolo_mark. There is a .txt file for each image with the same name. Each line in the label file is of the form: <object-class> <x> <y> <width> <height>.
Where:
<object-class>: integer number of object. (0) door; (1) handle; (2) cabinet door; (3) refrigerator door.<x> <y> <width> <height>: float values relative to width and height of the image.<x> <y>: center of the box.
The dataset can be used for training and testing an object detection CNN such as YOLO. Weights for detecting doors and handles with YOLO can be downloaded from: YOLO_weights (mAP=45%). For running YOLO you might also need the network configuration file yolo-obj.cfg and a text file where the detected classes names and their order is specified obj.names.
Please cite the paper in your publications if it helps your research.
@article{Arduengo_2021,
title={Robust and adaptive door operation with a mobile robot},
ISSN={1861-2784},
url={http://dx.doi.org/10.1007/s11370-021-00366-7},
DOI={10.1007/s11370-021-00366-7},
journal={Intelligent Service Robotics},
publisher={Springer Science and Business Media LLC},
author={Arduengo, Miguel and Torras, Carme and Sentis, Luis},
year={2021},
month={May}
}
Link to the paper: Robust and Adaptive Door Operation with a Mobile Robot
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent