Veterinary Data Quality Policy
Quality expectations for cattle health, disease, feed, and veterinary datasets.
Veterinary Data Quality Policy
This policy explains the quality standards required for animal images, diagnostic images, clinical information, and related data submitted to the platform.
Purpose of Data Quality Review
The platform reviews submitted data to ensure that animal images and related information are useful, accurate, original, relevant, and suitable for animal health datasets, veterinary research, AI model training, validation, testing, and product development.
Clear and Relevant Images
Uploaded images should clearly show the animal, body part, feed sample, disease condition, diagnostic image, or clinical finding related to the selected category. Images that do not clearly show the subject may be rejected.
Image Focus and Sharpness
Images should be focused and not excessively blurred. The affected area, lesion, wound, eye, hoof, udder, skin, feed material, diagnostic scan, or other submitted subject should be clearly visible.
Lighting and Exposure
Images should have proper lighting. Images that are too dark, too bright, shadowed, overexposed, underexposed, or unclear may not be useful for veterinary review or AI training and may be rejected.
Proper Framing
The important animal body region or diagnostic area should be properly framed. The image should not cut off the affected area or include too much irrelevant background that makes review difficult.
Correct Category Selection
Users must select the correct category while submitting data. For example, hoof images should be submitted under hoof-related categories, eye images under eye-related categories, skin lesions under skin-related categories, and feed images under feed-related categories.
Required Animal Information
Where available, users should provide important animal details such as species, breed, age, sex, body region, symptoms, disease condition, suspected diagnosis, confirmed diagnosis, severity, treatment details, and follow-up information.
Diagnosis Accuracy
Users should provide diagnosis information honestly. If the diagnosis is not confirmed, users should mark it as suspected, unknown, or not confirmed instead of submitting false or misleading information.
Veterinary Verification
Submitted data may be reviewed by authorized veterinary reviewers, platform administrators, data quality reviewers, AI teams, or research partners. Reviewers may approve, reject, correct, label, categorize, or request more information for a submission.
Image Originality
All submitted images must be real and original. Images copied from Google, websites, books, journals, social media, other apps, textbooks, or third-party sources without permission are not allowed and may be rejected.
Duplicate Detection
Users must not upload the same image multiple times or submit cropped, edited, resized, screenshot, or repeated versions of the same image to increase rewards. Duplicate and near-duplicate submissions may be rejected.
Fake or AI-Generated Images
Fake images, AI-generated animal images, manipulated images, staged images, misleading images, or images created only to earn rewards are strictly not allowed. Such submissions may lead to rejection, reward removal, or account suspension.
Privacy and Identifier Removal
Images should not contain owner names, phone numbers, addresses, human faces, Aadhaar or PAN details, prescription slips, clinic bills, private clinic records, exact GPS details, or other personal identifiers. Such data may be rejected or anonymized.
Clinical Usefulness
Images and information should be useful for veterinary review, animal health understanding, dataset development, or AI model improvement. Images that are unrelated, unclear, incomplete, or clinically meaningless may be rejected.
Minimum Metadata Quality
Submissions with missing or very limited information may be considered low quality. Animal species, image category, body region, symptoms or reason for upload, and diagnosis status should be provided wherever possible.
Severity and Case Details
Where relevant, users should provide disease severity, duration of symptoms, treatment already given, recovery status, and follow-up information. This improves the value of the data for veterinary research and AI development.
Diagnostic Image Quality
For X-rays, ultrasound images, pathology images, lab images, or other diagnostic images, the uploaded file should be readable, properly oriented, relevant to the selected case, and free from unnecessary personal identifiers.
Feed Image Quality
Feed images should clearly show the feed, fodder, ingredient, contamination, spoilage, foreign material, or quality issue being reported. Blurry, distant, or unclear feed images may be rejected.
Disease Image Quality
Disease images should clearly show the affected region, lesion, swelling, wound, discharge, skin change, hoof issue, eye issue, udder issue, or other visible clinical sign. The image should help reviewers understand the condition.
Data Consistency
Submitted information should match the uploaded image. For example, an image labelled as eye infection should show an eye-related issue, and an image labelled as hoof disease should show the hoof or limb region.
Low-Quality Submission Rejection
The platform may reject submissions that are blurry, irrelevant, duplicated, copied, fake, AI-generated, misleading, incorrectly labelled, incomplete, corrupted, privacy-risky, or not useful for veterinary review or AI training.
Reviewer Corrections
The platform may correct labels, categories, diagnosis status, severity, or metadata when reviewers identify errors. These corrections are made to improve dataset quality and reduce misleading information.
Data Exclusion from AI Training
Not all approved uploads may be used for AI training. Some images may be stored for review but excluded from AI datasets if they do not meet quality, balance, label confidence, privacy, or technical requirements.
Contributor Quality Score
The platform may monitor contributor quality based on approval rate, duplicate rate, rejection rate, metadata completeness, image usefulness, and policy violations. Poor-quality or fraudulent contributors may face reward restrictions or account suspension.
Reward Link to Data Quality
Rewards may depend on image quality, originality, metadata completeness, relevance, approval status, and usefulness for animal health datasets. Low-quality or incomplete submissions may not be eligible for rewards.
Continuous Data Improvement
The platform may update data quality standards from time to time to improve veterinary review, dataset reliability, AI model performance, bias control, and animal health technology development.
Contact Us
For questions about data quality, rejected submissions, correction requests, or upload guidelines, contact Chimertech Private Limited at research@chimertech.com or +91 97909 29442.