Polygonal Object Dataset

Datasets for Polygonal Object Detection

We propose two datasets to evaluate polygon object detection approaches: rooftop dataset and selected ImageNet categories. We manually labeled the outlines of the objects in each image as the ground truth.

All the ground truth outlines available in this page are copyrighted by CVLab-EPFL. You are free to use them for research purposes. If you use them to publish results, please cite the reference below.

Rooftop Dataset

This dataset consists of 65 aerial images of rural scenes containing several building rooftops many of which exhibit a fairly complex polygonal geometry. Each image is of size 1000 * 750 pixels.

Rooftop images

ImageNet Categories

This dataset includes images from 10 different object categories from ImageNet. They are sign, screen, remote control, cleaver, computer mouse, ipod, wine bottle, mug, beer bottle, and lampshade. We selected around 100 images for each category, and randomly split them into equal-sized training and testing sets.

We do not have coypright to publish the original images, we only provide the ground truth object outlines for corresponding images. To use this dataset, you have to download the input images from ImageNet website. If you use our ground truth outlines, please cite both ImageNet and the reference below.

ImageNet images

Reference

X. Sun; C. M. Christoudias; P. Fua : Free-Shape Polygonal Object Localization. 2014. European Conference on Computer Vision (ECCV), Zurich, Switzerland, Sep 6-12, 2014. p. 317-332.