HOWA dataset (v1) - HOusehold WAste dataset

This dataset contains scenes with household waste objects on the soil. It is composed of color images in several terrestrial environments and it can be used to train and evaluate waste detection and segmentation methods. Each image contains one target in different conditions, including variance of depth, illumination, occlusion and object erosion. There are 6943 images labeled according to the following four classes: plastic, carton, glass and metal.

The next table shows the sample distribution for each class:

Class #samples Description
Plastic 1776 Objects made of plastic.
Carton 1626 Objects made of carton.
Glass 1769 Objects made of glass.
Metal 1772 Objects made of metal.

For evaluation we have provided the dataset separated in training, validation and testing sets, following the 70/20/10 proportion, resulting in the following distribution:

Class #samples train set #samples validation set #samples test set
Plastic 1240 350 186
Carton 1114 339 173
Glass 1263 334 171
Metal 1243 365 164

All environments are in the University of Alicante area and include asphalt, pebbles and green backgrounds.

1 2 3 4
1 2 3 4

LICENSE AND ACCESS

This dataset is shared only for non-profit research or educational purposes. If you use this dataset or a part of it, please respect these terms of use and reference the original work in which it was published.

FILE FORMAT

The images were acquired with a RealSense™ D435i camera. The size of the images are 640 x 480 pixels. The images are stored as JPG where pixel values represent RGB colors. The distance between objects and camera has been changed in order to obtain closer and further images.

All the images are distributed in train / validation / test sets following the aforementioned 70 / 20 / 10 distribution.

In addition, the bounding box and segmentation area of each object is included. This labeling is stored in a JSON file, one for each image. For being able to convert this dataset to COCO format, we need to use the tool labelme2coco.py, which is inside the labelme repository.

RELATED PUBLICATIONS (CITATION)

Please, if you use this dataset or part of it, cite the following publication: 


  
  
  @article{Paez-Ubieta_Castano-Amoros_Puente_Gil_2023, 
  title={Vision and tactile robotic system to grasp litter in outdoor environments},
  volume={109}, 
  DOI={10.1007/s10846-023-01930-2}, 
  number={2},
  journal={Journal of Intelligent & Robotic Systems}, 
  author={Páez-Ubieta, Ignacio De Loyola and Castaño-Amorós, Julio and Puente, Santiago T. and Gil, Pablo}, 
  year={2023}, 
  month={Oct}
  }

This work was funded by the Spanish Government through the project RTI2018-094279-B-I00 and the Spanish Valencian Government through the projects IDIFEDER/2020/003 and PROMETEO/2021/075.

DOWNLOAD

To download this dataset fill in the following form.

VERSIONS

V1 - Released in February 2023. This version of the dataset contained 6943 images separated in train, validation and test sets. The dataset has a JSON file for each image.