Our Projects

Research We've
Powered.

Real annotations by licensed veterinary experts. Every number below reflects data we labeled.

KeypointRadiologyDog

Canine Hip Dysplasia Detection

2,306 radiographs annotated with keypoints for AI-powered Norberg angle measurement.

Pakistan Vet Journal · 2025

View details →

Bounding BoxMulti-SpeciesForensic

Smart Osteology

26,148 bone images across 3 species labeled for two-stage detection and classification.

Veterinary Sciences · 2025

View details →

Object DetectionReproductionBull

Bull Sperm Viability Assessment

3,068 microscopy images with live/dead sperm annotations, benchmarked across 4 models.

Frontiers in Vet Science · 2026

View details →

Bounding BoxWildlife9 Species

Human-Wildlife Interface Detection

Camera trap imagery for 9 wild species near Aladaglar National Park, targeting road safety.

TÜBİTAK · Ongoing

View details →

Keypoint Annotation · Radiology · Dog

Canine Hip Dysplasia Detection

Ventro-dorsal pelvis radiographs annotated with bounding boxes and 4 anatomical keypoints per image, enabling AI to automatically measure the Norberg angle and diagnose hip dysplasia.

Data We Labeled

2,306

Radiographs

DICOM format, dogs 12 mo+

Keypoints / Image

Acetabulum + femur head L/R

9,224+

Total Keypoints

Sub-pixel accuracy

0.994

mAP@50

YOLOv8 best config

Annotation Details

Each image labeled twice: hip region bounding box + 4 anatomical keypoints

Landmarks: center of right/left caput femoris and dorsocranial projection of right/left acetabulum

Labels validated by licensed veterinary surgeons against radiographic standards

All YOLOv8–v12 model families benchmarked on the labeled dataset

Published in Pakistan Veterinary Journal, 45(4):1930–1937, 2025.
📄 View Article →

Bounding Box · Multi-Species · Forensic

Smart Osteology

26,148 bone images across cattle, horse, and dog labeled with bounding boxes, feeding a two-stage YOLOv5 + CNN pipeline for bone type and species classification.

Data We Labeled

26,148

Images

2,744 per class combination

Classes

3 species × 3 bone types

460

Specimens

Cattle, horse, dog bones

97.6%

Accuracy

YOLOv5 + CNN pipeline

Annotation Details

Bounding boxes on scapula, humerus, and femur across 3 species

Balanced dataset: 2,744 images per class regardless of specimen count

Images captured at 720p from multiple angles and lighting conditions

Labels feed a two-stage pipeline: YOLOv5 detects, CNN classifies species and bone type

Published in Veterinary Sciences, 12(8):765, 2025.
📄 View Article →

Object Detection · Reproduction · Bull

Bull Sperm Viability Assessment

3,068 eosin/nigrosin-stained microscopy images annotated per-cell, classifying each spermatozoon as live or dead. Benchmarked across 4 architectures.

Data We Labeled

3,068

Photos

2,140 train / 512 val / 416 test

Classes

Live · Dead spermatozoa

97.1%

Accuracy

YOLOv8 balanced accuracy

0.978

F1 Score

YOLOv8 test set

Annotation Details

Each spermatozoon individually labeled as live or dead per WHO eosin/nigrosin criteria

15 frozen bull semen batches thawed under standardized lab conditions

Independent mock test: 281 sperm across 36 images, AI vs. expert — no significant difference

4 architectures compared: YOLOv8, YOLOv12, Faster R-CNN, YOLOv5 + HRNet

Published in Frontiers in Veterinary Science, 13:1772252, March 2026.
📄 View Article →

Bounding Box · Wildlife · 9 Species

Human-Wildlife Interface Detection

Camera trap and field imagery along the Kayseri-Nigde-Adana road near Aladaglar National Park, annotated for 9 wild species to reduce wildlife-traffic collisions.

Data We Labeled

Species

Mammal + avian classes

500+

Images/Species

Minimum target

1,142

Road Incidents

Animal collisions TR 2025

YOLOv12

Target Model

+ SAHI algorithm

Annotation Details

Species: mountain goat, wild horse, roe deer, wolf, fox, wild boar, eagle, hawk, owl

Sources: camera traps, MOBESE surveillance nodes, field photography

Annotation via Roboflow with generative AI augmentation for rare species

Final system integrates with road monitoring for real-time wildlife alerts

TÜBİTAK-supported conservation project — Team Arcuray, Erciyes University Veterinary Faculty.

Research We'vePowered.

Canine Hip Dysplasia Detection

Smart Osteology

Bull Sperm Viability Assessment

Human-Wildlife Interface Detection

Canine Hip Dysplasia Detection

Smart Osteology

Bull Sperm Viability Assessment

Human-Wildlife Interface Detection

Research We've
Powered.