<?xml version='1.0' encoding='UTF-8'?><codeBook xmlns="ddi:codebook:2_5" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="ddi:codebook:2_5 https://ddialliance.org/Specification/DDI-Codebook/2.5/XMLSchema/codebook.xsd" version="2.5"><docDscr><citation><titlStmt><titl>Supporting Data for: UltraMNIST Classification: A Benchmark to Train CNNs for Very Large Images</titl><IDNo agency="DOI">doi:10.18710/4F4KJS</IDNo></titlStmt><distStmt><distrbtr source="archive">DataverseNO</distrbtr><distDate>2023-03-29</distDate></distStmt><verStmt source="archive"><version date="2023-09-28" type="RELEASED">1</version></verStmt><biblCit>Gupta, Deepak K.; Bhamba, Udbhav; Thakur, Abhishek; Gupta, Akash; Sharan, Suraj; Demir, Ertugrul; Prasad, Dilip K., 2023, "Supporting Data for: UltraMNIST Classification: A Benchmark to Train CNNs for Very Large Images", https://doi.org/10.18710/4F4KJS, DataverseNO, V1</biblCit></citation></docDscr><stdyDscr><citation><titlStmt><titl>Supporting Data for: UltraMNIST Classification: A Benchmark to Train CNNs for Very Large Images</titl><IDNo agency="DOI">doi:10.18710/4F4KJS</IDNo></titlStmt><rspStmt><AuthEnty affiliation="UiT The Arctic University of Norway">Gupta, Deepak K.</AuthEnty><AuthEnty affiliation="UiT The Arctic University of Norway">Bhamba, Udbhav</AuthEnty><AuthEnty affiliation="UiT The Arctic University of Norway">Thakur, Abhishek</AuthEnty><AuthEnty affiliation="UiT The Arctic University of Norway">Gupta, Akash</AuthEnty><AuthEnty affiliation="UiT The Arctic University of Norway">Sharan, Suraj</AuthEnty><AuthEnty affiliation="UiT The Arctic University of Norway">Demir, Ertugrul</AuthEnty><AuthEnty affiliation="UiT The Arctic University of Norway">Prasad, Dilip K.</AuthEnty><othId role="Researcher">Gupta, Deepak K.</othId><othId role="Researcher">Bhamba, Udbhav</othId><othId role="Researcher">Thakur, Abhishek</othId><othId role="Researcher">Gupta, Akash</othId><othId role="Researcher">Sharan, Suraj</othId><othId role="Researcher">Demir, Ertugrul</othId><othId role="Researcher">Prasad, Dilip K.</othId><othId role="Other">Nirwan Banerjee</othId></rspStmt><prodStmt><producer abbr="UiT">UiT The Arctic University of Norway</producer><software version="3">Python</software><software version="3">Numpy</software><grantNo agency="The Research Council of Norway">grant no. 325741</grantNo><grantNo agency="UiT The Arctic University of Norway">Cristin project id 2061348</grantNo></prodStmt><distStmt><distrbtr source="archive">DataverseNO</distrbtr><distrbtr affiliation="UiT The Arctic University of Norway" URI="https://dataverse.no/dataverse/uit">UiT The Arctic University of Norway</distrbtr><contact affiliation="UiT The Arctic University of Norway" email="dilip.prasad@uit.no">Prasad, Dilip K.</contact><depositr>Banerjee, Nirwan</depositr><depDate>2023-03-06</depDate></distStmt><serStmt><serName>Ultra-MNIST</serName><serInfo>https://www.kaggle.com/competitions/ultra-mnist</serInfo></serStmt><holdings URI="https://doi.org/10.18710/4F4KJS"/></citation><stdyInfo><subject><keyword xml:lang="en">Computer and Information Science</keyword><keyword xml:lang="en">Medicine, Health and Life Sciences</keyword><keyword>Large Scale Image Classification</keyword><keyword>Ultra Large MNIST dataset</keyword><keyword>Variable scale features</keyword><keyword>CNN classification</keyword></subject><abstract date="2022-04-15">Convolutional neural network (CNN) approaches available in the current literature are designed to work primarily with low-resolution images. When applied on very large images, challenges related to GPU memory, smaller receptive field than needed for semantic correspondence and the need to incorporate multi-scale features arise. The resolution of input images can be reduced, however, with significant loss of critical information. Based on the outlined issues, we introduce a novel research problem of training CNN models for very large images, and present ‘UltraMNIST dataset’, a simple yet representative benchmark dataset for this task. UltraMNIST has been designed using the popular MNIST digits with additional levels of complexity added to replicate well the challenges of real-world problems. We present two variants of the problem: ‘UltraMNIST classification’ and ‘Budget-aware UltraMNIST classification’. The standard UltraMNIST classification benchmark is intended to facilitate the development of novel CNN training methods that make the effective use of the best available GPU resources. The budget-aware variant is intended to promote development of methods that work under constrained GPU memory. For the development of competitive solutions, we present several baseline models for the standard benchmark and its budget-aware variant. We study the effect of reducing resolution on the performance and present results for baseline models involving pretrained backbones from among the popular state-of-the-art models. Finally, with the presented benchmark dataset and the baselines, we hope to pave the ground for a new generation of CNN methods suitable for handling large images in an efficient and resource-light manner. UltraMNIST dataset comprises very large-scale images, each of 4000x4000 pixels with 3-5 digits per image. Each of these digits has been extracted from the original MNIST dataset. Your task is to predict the sum of the digits per image, and this number can be anything from 0 to 27.</abstract><sumDscr><collDate cycle="P1" event="start" date="2022-04-15">2022-04-15</collDate><dataKind>Image data</dataKind><dataKind>Large Scale data</dataKind><dataKind>Variable Scale Feature data</dataKind></sumDscr></stdyInfo><method><dataColl><sources><dataSrc>https://www.kaggle.com/competitions/ultra-mnist</dataSrc></sources></dataColl><anlyInfo/></method><dataAccs><setAvail/><useStmt/><notes type="DVN:TOU" level="dv">&lt;a href="http://creativecommons.org/publicdomain/zero/1.0">CC0 1.0&lt;/a></notes></dataAccs><othrStdyMat><relPubl><citation><titlStmt><titl>Gupta, D. K., Bamba, U., Thakur, A., Gupta, A., Sharan, S., Demir, E., &amp; Prasad, D. K. (2022). UltraMNIST Classification: A Benchmark to Train CNNs for Very Large Images.</titl><IDNo agency="doi">10.48550/arXiv.2206.12681</IDNo></titlStmt><biblCit>Gupta, D. K., Bamba, U., Thakur, A., Gupta, A., Sharan, S., Demir, E., &amp; Prasad, D. K. (2022). UltraMNIST Classification: A Benchmark to Train CNNs for Very Large Images.</biblCit></citation><ExtLink URI="https://doi.org/10.48550/arXiv.2206.12681"/></relPubl></othrStdyMat></stdyDscr><otherMat ID="f176287" URI="https://dataverse.no/api/access/datafile/176287" level="datafile"><labl>00_ReadMe.txt</labl><notes level="file" type="DATAVERSE:CONTENTTYPE" subject="Content/MIME Type">text/plain</notes></otherMat><otherMat ID="f175754" URI="https://dataverse.no/api/access/datafile/175754" level="datafile"><labl>test.zip</labl><notes level="file" type="DATAVERSE:CONTENTTYPE" subject="Content/MIME Type">application/zip</notes></otherMat><otherMat ID="f175731" URI="https://dataverse.no/api/access/datafile/175731" level="datafile"><labl>train.csv</labl><txt>For testing purposes you may contact Dilip K. Prasad at dilip.prasad@uit.no</txt><notes level="file" type="DATAVERSE:CONTENTTYPE" subject="Content/MIME Type">text/csv</notes></otherMat><otherMat ID="f175752" URI="https://dataverse.no/api/access/datafile/175752" level="datafile"><labl>train.zip</labl><txt>The training files containing all the images of the training set.</txt><notes level="file" type="DATAVERSE:CONTENTTYPE" subject="Content/MIME Type">application/zip</notes></otherMat></codeBook>