2018 Hoarding and INDoor cluttER (HINDER) Dataset

Motivation

Collecting and rating images of hoarding-related indoor clutter is difficult and labor-intensive. The images need to be rated according to the Clutter Image Rating (CIR) scale from 1 (no clutter) to 9 (highest clutter)  [R.O. Frost, G. Steketee. D.F. Tolin and S. Renaud], which is labor-intensive and often ambiguous. Even trained professionals admit assigning CIR values within ±1. The original 2018 Hoarding and INDoor cluttER (HINDER) dataset described in [Tezcan et al., 2018]  includes a mixture of training (1,233) and validation (90) images for the total of 1,323 images. The CIR class membership in this dataset is highly unbalanced as shown in the table below.

Description

HINDER-2018 includes 1,323 images of different sizes, aspect ratios and visual appearance (lighting, contrast, compression artifacts). All images are in JPEG format, grouped into 9 folders (separate folder for each CIR class). The filename of each image includes the CIR class and a random hash string. The downloadable file is a single ZIP file containing all folders and images.

Dataset Download

You may use this dataset for non-commercial purposes. If you publish any work reporting results using this dataset, please cite the following paper:

M. Tezcan, J. Konrad, and J. Muroff, “Automatic assessment of hoarding clutter from images using convolutional neural networks,” in Proc. IEEE Southwest Symposium on Image Analysis and Interpretation, Apr. 2018.

To access the download page, please complete the form below (tested only in Chrome).

Contact

Please contact [jkonrad] at [bu] dot [edu] if you have any questions.

Acknowledgements

We would also like to thank Boston University students for their help in collecting and rating images in this dataset.