Panoramic Object Detection Dataset

Contents

This repository contains following contents:

• the dataset collected and used in "Object Detection in Equirectangular Panorama"
• the evaluation code (to be added)
• the comparison experiment results
• the annotation tool

Paper

Object Detection in Equirectangular Panorama

Wenyan Yang^*, Yanlin Qian^*, Joni-Kristian Kämäräinen^*, Francesco Cricri^*, Lixin Fan^*

International Conference on Pattern Recognition (ICPR) 2018

Dataset

Downloads: link Slides: link

It containes three types of the ground truth: BFOV (Bounding FOV), BBox(Bounding box) and 3D-Container frame.

• BFOV (path: 360-Dataset/bbx-annotations/*)

<bndbox>
	<xmin>
	<ymin>
	<xmax>
	<ymax>
</BBOX> 

• BFOV (path: 360-Dataset/bfov-annotations/*)

<BFOV> 
	<center_lon> // centre's longitude
	<center_lat> // centre's latitude
	<bfov_w> // the width's radius
	<bfov_h> //  the height's radius
</BFOV> 

• 3D-Container Frame(path: 360-Dataset/bfov-annotations/*)

<ThreeDbox> 
// The 4 vertexes and 1 centre point of the BFOV. Stored in radians
      <tl_lon> // top left vertex's longitude.
      <tl_lat> // top left vertex's latitude.
      <tr_lon> // top right vertex's longitude.
      <tr_lat> // top right vertex's latitude.
      <br_lon> // bottom right vertex's longitude.
      <br_lat> // bottom right vertex's latitude.
      <bl_lon> // bottom left vertex's longitude.
      <bl_lat> // bottom left vertex's latitude.
      <c_lon> // centre's longitude
      <c_lat> // centre's latitude.
</ThreeDbox>

Visualization

Youtube: link

• To visualize the BBOX annotation, some traditional annotation tools are needed (e.g: https://github.com/tzutalin/labelImg)
• To visualize the BFOV annotation, use the tool in folder "/visualize_tool" (open the web browser and load the image and XML file)

Annotation tool

To be added

Comparison experiments results

To be added

Running the tests

To be added

License

This project is licensed under the MIT License - see the LICENSE.md file for details

Citation

• Please cite our paper if you find this work is useful.

@INPROCEEDINGS{8546070, 
author={W. Yang and Y. Qian and J. Kämäräinen and F. Cricri and L. Fan}, 
	booktitle={2018 24th International Conference on Pattern Recognition (ICPR)}, 
	title={Object Detection in Equirectangular Panorama}, 
	year={2018}
	pages={2190-2195}, 
	keywords={Detectors;Distortion;Object detection;Cameras;Image resolution;Virtual reality;Graphics processing units}, 
	doi={10.1109/ICPR.2018.8546070}, 
	ISSN={1051-4651}, 
	month={Aug}
}