October 20, 2022 Dataset Open Access

Knauer, Markus; Denninger, Maximilian; Triebel, Rudolph

DataCite XML Export

<?xml version='1.0' encoding='utf-8'?>
<resource xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns="http://datacite.org/schema/kernel-4" xsi:schemaLocation="http://datacite.org/schema/kernel-4 http://schema.datacite.org/meta/kernel-4.1/metadata.xsd">
  <identifier identifierType="DOI">10.5281/zenodo.7189434</identifier>
  <creators>
    <creator>
      <creatorName>Knauer, Markus</creatorName>
      <givenName>Markus</givenName>
      <familyName>Knauer</familyName>
      <nameIdentifier nameIdentifierScheme="ORCID" schemeURI="http://orcid.org/">0000-0001-8229-9410</nameIdentifier>
      <affiliation>German Aerospace Center (DLR)</affiliation>
    </creator>
    <creator>
      <creatorName>Denninger, Maximilian</creatorName>
      <givenName>Maximilian</givenName>
      <familyName>Denninger</familyName>
      <nameIdentifier nameIdentifierScheme="ORCID" schemeURI="http://orcid.org/">0000-0002-1557-2234</nameIdentifier>
      <affiliation>German Aerospace Center (DLR), Technical University of Munich (TUM)</affiliation>
    </creator>
    <creator>
      <creatorName>Triebel, Rudolph</creatorName>
      <givenName>Rudolph</givenName>
      <familyName>Triebel</familyName>
      <nameIdentifier nameIdentifierScheme="ORCID" schemeURI="http://orcid.org/">0000-0002-7975-036X</nameIdentifier>
      <affiliation>German Aerospace Center (DLR), Technical University of Munich (TUM)</affiliation>
    </creator>
  </creators>
  <titles>
    <title>HOWS-CL-25: Household Objects Within Simulation Dataset for Continual Learning</title>
  </titles>
  <publisher>Zenodo</publisher>
  <publicationYear>2022</publicationYear>
  <subjects>
    <subject>object classification</subject>
    <subject>instance segmentation</subject>
    <subject>depth estimation</subject>
    <subject>household objects</subject>
    <subject>continual learning</subject>
    <subject>online learning</subject>
    <subject>incremental learning</subject>
    <subject>mobile robot</subject>
    <subject>BlenderProc</subject>
    <subject>Blender</subject>
    <subject>synthetic images</subject>
    <subject>RGB images</subject>
    <subject>segmentation images</subject>
    <subject>normal images</subject>
    <subject>depth images</subject>
    <subject>neural network</subject>
    <subject>CNN</subject>
    <subject>Convolutional neural network</subject>
    <subject>Learning</subject>
    <subject>Cognitive Intelligence CI</subject>
  </subjects>
  <dates>
    <date dateType="Issued">2022-10-20</date>
  </dates>
  <language>en</language>
  <resourceType resourceTypeGeneral="Dataset"/>
  <alternateIdentifiers>
    <alternateIdentifier alternateIdentifierType="url">https://zenodo.org/record/7189434</alternateIdentifier>
  </alternateIdentifiers>
  <relatedIdentifiers>
    <relatedIdentifier relatedIdentifierType="DOI" relationType="IsVersionOf">10.5281/zenodo.7054170</relatedIdentifier>
  </relatedIdentifiers>
  <version>1.0.1</version>
  <rightsList>
    <rights rightsURI="https://creativecommons.org/licenses/by/4.0/legalcode">Creative Commons Attribution 4.0 International</rights>
    <rights rightsURI="info:eu-repo/semantics/openAccess">Open Access</rights>
  </rightsList>
  <descriptions>
    <description descriptionType="Abstract">&lt;p&gt;HOWS-CL-25 (Household Objects Within Simulation dataset for Continual Learning) is a synthetic dataset especially designed for object classification on mobile robots operating in a changing environment (like a household), where it is important to learn new, never seen objects on the fly.&lt;br&gt;
This dataset can also be used for other learning use-cases, like instance segmentation or depth estimation.&lt;br&gt;
Or where household objects or continual learning are of interest.&lt;/p&gt;

&lt;p&gt;Our dataset contains 150,795 unique synthetic images using 25 different household categories with 925 3D models in total. For each of those categories, we generated about 6000 RGB images. In addition, we also provide a corresponding depth, segmentation, and normal image.&lt;/p&gt;

&lt;p&gt;The dataset was created with BlenderProc [Denninger et al. (2019)], a procedural pipeline to generate images for deep learning.&lt;br&gt;
This tool created a virtual room with randomly textured floors, walls, and a light source with randomly chosen light intensity and color. After that, a 3D model is placed in the resulting room. This object gets customized by randomly assigning materials, including different textures, to achieve a diverse dataset. Moreover, each object might be deformed with a random&lt;br&gt;
displacement texture.&lt;br&gt;
We use 774 3D models from the ShapeNet dataset [A. X. Chang et al. (2015)] and the other models from various internet sites. Please note that we had to manually fix and filter most of the models with Blender before using them in the pipeline!&lt;/p&gt;

&lt;p&gt;For continual learning (CL), we provide two different loading schemes:&lt;br&gt;
- Five sequences with five categories each&lt;br&gt;
- Twelve sequences with three categories in the first and two in the other sequences.&lt;/p&gt;

&lt;p&gt;In addition to the RGB, depth, segmentation, and normal images, we also provide the calculated features of the RGB images (by ResNet50) as used in our RECALL paper.&lt;br&gt;
In those two loading schemes, ten percent of the images are used for validation, where we ensure that an object instance is either in the training or the validation set, not in both. This&amp;nbsp;avoids learning&amp;nbsp;to recognize certain instances by heart.&lt;/p&gt;

&lt;p&gt;We recommend using those loading schemes to compare your approach with others.&lt;/p&gt;

&lt;p&gt;Here we provide three files for download:&lt;br&gt;
- HOWS_CL_25.zip [124GB]: This is the original dataset with the RGB, depth, segmentation, and normal images, as well as the loading schemes. It is divided into three archive parts. To open the dataset, please ensure to download all three parts.&lt;br&gt;
- HOWS_CL_25_hdf5_features.zip [2.5GB]: This only contains the calculated features from the RGB input by a ResNet50 in a .hdf5 file. Download this if you want to use the dataset for learning and/or want to compare your approach to our RECALL approach (where we used the same features).&lt;br&gt;
- README.md: Some additional explanation.&lt;/p&gt;

&lt;p&gt;For further information and code examples, please have a look at our website: https://github.com/DLR-RM/RECALL.&lt;/p&gt;</description>
    <description descriptionType="Other">{"references": ["M. Denninger, M. Sundermeyer, D. Winkelbauer, D. Olefir, T. Hodan, \u02c7 Y. Zidan, M. Elbadrawy, M. Knauer, H. Katam, and A. Lodhi, \"Blenderproc: Reducing the reality gap with photorealisitc rendering,\" Robotics: Science and Systems (RSS), 2020", "A. X. Chang, T. Funkhouser, L. Guibas, P. Hanrahan, Q. Huang, Z. Li, S. Savarese, M. Savva, S. Song, H. Su et al., \"Shapenet: An information-rich 3d model repository,\" arXiv:1512.03012, 2015."]}</description>
  </descriptions>
</resource>