What is included in the dataset?
It contains 16,066 image-based annotation records collected from Q1 2022 through Q2 2024, plus petType, breed, petSex, petSize, petHair, neuterYn, petBirth, poopColor, and poopShape metadata.
Pet dataset for AI and computer vision research
16,066 labeled pet stool condition images
A B2B dataset for research institutes, university labs, and pet health R&D teams. Review the data structure, field schema, and per-field coverage first, then ask about a license that fits your research.
This dataset is for research, education, ML development, and service evaluation. Coordinates marking the stool region inside each image (bounding boxes) are not included.
Buy now · Ask about purchase or licensing
Representative records, published as they are
Representative data preview
Data collected through the Barabom app
When users record pet stool entries in the Barabom app, images and condition labels are collected together. The video below shows the actual recording flow.
The video shows the collection flow. Scope and licensing terms are discussed separately.
Ways to use pet image data for AI training
Use dog and cat stool images with label data for computer vision model training, image classification research, and AI image training data review. The source images and metadata are suitable for research institutes, university labs, pet health R&D teams, and data buyers to review quality.
- AI research at institutes and universities - Research institutes, universities, and graduate labs can use the data to train and validate computer vision models.
- Graduate theses and industry academic projects - Use the image labeling data for cross-disciplinary research in AI, data science, and veterinary medicine.
- Computer vision image classification - Train and compare models that classify color and shape labels in pet stool images.
- Pet health R&D and product PoCs - Pet healthcare companies and veterinary R&D teams can use the data for observation support service PoCs.
- Label validation for AI and data teams - Review preprocessing, label consistency, and error analysis for AI training image datasets.
- AI data sourcing and procurement review - Companies, institutions, and data purchasing teams can review source data and labeled data.
Quarterly records and coverage
Record counts are based on the labeled records in each year and quarter. Coverage is the share of records holding a real value once missing and empty values are excluded. Because owners enter the data themselves it may differ from the actual condition, so missing counts are published as they are.
Quarterly annotation records
- 16,066 Labeled images - Annotation records collected from Q1 2022 through Q2 2024.
- 9 fields Metadata per record - petType, breed, petSex, petSize, petHair, neuterYn, petBirth, poopColor, poopShape
- 99.80% Highest field coverage - Measured on neuterYn. Coverage and missing counts are published as they are.
- 2022 1Q: 99 records, 2022 2Q: 546 records, 2022 3Q: 542 records, 2022 4Q: 774 records
- 2023 1Q: 1,387 records, 2023 2Q: 1,845 records, 2023 3Q: 2,629 records, 2023 4Q: 2,675 records
- 2024 1Q: 2,647 records, 2024 2Q: 2,922 records
Field-level coverage
- Birth date (petBirth): 99.43% - 15,975 records / missing 91. Records where petBirth contains a real date string
- Body size (petSize): 87.41% - 14,044 records / missing 2,022. Records entered as small, medium, or large
- Hair length (petHair): 92.80% - 14,910 records / missing 1,156. Records entered as l or s
- Neutered (neuterYn): 99.80% - 16,034 records / missing 32. Records entered as y or n
- Stool color (poopColor): 81.71% - 13,128 records / missing 2,938. Records with a non-null color label
- Stool shape (poopShape): 96.79% - 15,550 records / missing 516. Records with a non-null shape or condition label
Actual enum value distribution
Animal type (petType)
Four values, dogs and cats included
- dog: 62.91% - 10,107 records
- cat: 36.99% - 5,943 records
- small_animal: 0.05% - 8 records
- other: 0.05% - 8 records
Sex (petSex)
Two values
- f: 52.74% - 8,474 records
- m: 47.26% - 7,592 records
Body size (petSize)
Small, medium, large and missing share
- small: 50.20% - 8,065 records
- medium: 26.55% - 4,266 records
- large: 10.66% - 1,713 records
- null: 12.59% - 2,022 records
Hair length (petHair)
Long, short and missing share
- l: 49.50% - 7,953 records
- s: 43.30% - 6,957 records
- null: 7.20% - 1,156 records
Neutered (neuterYn)
Neutered status and missing share
- y: 67.29% - 10,811 records
- n: 32.51% - 5,223 records
- null: 0.20% - 32 records
Stool color (poopColor)
Seven colour labels
- Chocolate: 75.48% - 12,126 records
- null: 18.29% - 2,938 records
- Yellow: 3.08% - 495 records
- Black: 1.52% - 245 records
- Bloody: 1.18% - 189 records
- Green: 0.43% - 69 records
- Purple: 0.02% - 4 records
Stool shape (poopShape)
Eight shape labels
- NORMAL_1: 64.45% - 10,354 records
- DIARRHEA_2: 10.18% - 1,635 records
- NORMAL_2: 8.65% - 1,389 records
- DIARRHEA_1: 5.52% - 887 records
- DRY_1: 3.53% - 567 records
- null: 3.21% - 516 records
- DIARRHEA_4: 2.88% - 462 records
- DIARRHEA_3: 1.59% - 256 records
Key purchase information
Frequently asked questions about the stool image dataset
Clear answers about dataset size, included fields, price, license scope, and delivery.
Page last reviewed:
How many dog and cat records are included?
The petType field contains 10,107 dog records, 5,943 cat records, 8 small_animal records, and 8 other records, for 16,066 records in total.
How much does the dataset cost?
The standard purchase price is KRW 3,900,000. Contact Barabom before purchase if you need a separate commercial partnership or extended license.
What does the license allow?
Buyers may use the data for internal research, analysis, AI model development, and service evaluation. Raw-data resale, unauthorized redistribution, and uploads to public repositories are prohibited.
How is the dataset delivered after payment?
After payment is verified, a download link is sent to the email address entered at checkout. The link is valid for 30 days and supports up to 10 downloads.
What limitations should teams review before training?
The data was submitted by Barabom app users and may contain label inconsistencies, missing values, and image-quality variation. Coordinates marking the stool region inside each image (bounding boxes) are not included either.
Review the data, then ask about a license
Tell us your research goal and intended scope of use, and we will confirm what we can provide.
Business information
- Business name: Barabom
- Representative: KimJeonghun
- Business registration number: 460-08-02680
- Business address: 50 Haedeung-ro, Dobong-gu, Seoul, Republic of Korea
- Business type/item: Information and communication / Application software publishing
- Mail-order business report number: 제 2026-서울도봉-0268 호
- Customer center: +82 10-2286-9185