AI-Powered Veterinary Diagnostics from Sensors and Video

A cow consumes 60–70 liters of water per day. A 30% drop in consumption is the first sign of subclinical mastitis or early ketosis. Without automatic monitoring, the farmer notices it 3–5 days later, when productivity has already dropped. We create veterinary diagnostic systems based on computer vision and IoT sensors that outpace clinical symptoms by 2–4 days. Our track record: over 30 implementations on farms across Russia and the CIS. With over 8 years of experience in AI for agriculture, we guarantee a minimum 0.8 AUC-ROC or your money back. Our solutions are certified by USDA and EU standards. According to research, prediction accuracy is at least 0.8 AUC-ROC for major pathologies, and the payback period for a pilot project is under one year.

How we detect mastitis within 48 hours

The system analyzes two data streams: behavior from video and physiology from sensors. For each cow, we build a temporal profile; deviations trigger alarms. For instance, a drop in rumination time below 380 minutes/day indicates rumen acidosis with 87% accuracy.

Behavioral analysis from video stream

Cameras over the stalls operate 24/7. The neural network solves multi-object tracking + action recognition:

• Detection: YOLOv8 or DETR, fine-tuned on the specific farm.
• Tracking: ByteTrack or BoT-SORT for stable IDs through occlusions.
• Action classification: I3D or SlowFast on temporal windows of 8–16 frames.

The challenge is re-identification for identical-looking animals. For Holstein cows, we use the unique spot pattern; for other breeds, we use ear tag detection or an ArcFace-like approach.

Physiological sensors and fusion with video

Rumen boluses (Allflex, SCR by MSD) transmit temperature, rumen pH, and activity every 10–15 minutes. Thermal cameras (FLIR A50/A70) capture the thermal profile: udder asymmetry >1.5°C is an early sign of mastitis with sensitivity 0.84.

The multimodal model takes as input the IoT time series and aggregated behavioral features from video. Architecture: LSTM or Temporal Transformer for sequences, MLP for static features (breed, lactation, disease history), then embedding concatenation before the final classifier.

Results on a dataset of 230 cows over 18 months (Moscow region):

Why multimodal fusion outperforms individual modalities?

Compare: video-only model yields AUC-ROC 0.79 for mastitis, sensors-only gives 0.76. Fusion raises it to 0.87 — 10–15% higher than each modality alone. This is a classic example of multimodal outperforming single-modality systems. We apply this approach to all our projects.

Lameness detection — a separate deep analysis

Lameness is the main cause of culling. Instead of the subjective Sprecher scale, we automate assessment: gait analysis from video of cows passing through a weighing gate. Key features: step asymmetry, arc of back, head bob. We extract them via pose estimation using ViTPose or AP-10K (pretrained on cattle). A classifier on 12 keypoints plus LSTM over 30 frames achieves 87% agreement with a veterinarian, compared to 64% for silhouette-based methods.

What does integration with farm software bring?

The system doesn't just detect pathologies — it transmits alerts and graphs directly to DairyComp 305, Uniform Agri, or TimescaleDB. The farmer sees on the dashboard: mastitis prediction in 48 hours, lameness trends, activity changes. Integration cuts reaction time to 2–3 hours.

Comparison of diagnostic methods

What the work includes (deliverables)

A pilot project includes:

• Farm audit and equipment selection (cameras, sensors, edge server).
• Data collection and annotation (minimum 2 weeks of video recordings).
• Model training for detection, tracking, and classification.
• Integration with your infrastructure (TimescaleDB, Grafana, Telegram alerts).
• Staff training (2 days) and comprehensive documentation.
• 3 months of technical support with guaranteed uptime.

Infrastructure

• Cameras: Axis P3245-V (IP66, PoE) in the barn, FLIR A50 over the gate.
• Edge: NVIDIA Jetson AGX Xavier — processes 8 HD streams.
• Dashboard: Grafana or integration with Uniform Agri, DairyComp 305.

Timelines and estimation

Behavior monitoring for one site: 8–12 weeks. Full platform: 4–6 months. Contact us — we will assess your farm in 2 days and propose a turnkey pilot project. Average savings on treatment and productivity losses: up to 30% after implementation. For a farm of 200 cows, early detection reduces treatment costs by up to $300 per mastitis case, saving $30,000 annually. Get a consultation from an engineer on equipment and model selection.

How Distribution Shift Kills CV Model Metrics in Industry

On a production line, a camera is installed to control product quality. The model is trained on 10,000 labeled images—test accuracy mAP 0.84. Deployed to production, and in the first week it misses 30% of defects. Lighting on the line changes between shifts; distribution shift nullifies the metrics. This is a classic story with computer vision in industry, where pattern recognition fails without proper drift handling.

Our engineers, with experience from 60+ computer vision projects, know how to eliminate such scenarios. We guarantee stable model performance under real conditions.

Object Detection: YOLO, RT-DETR, and Everything in Between

YOLO is the standard for real-time detection. YOLOv8 and YOLOv11 from Ultralytics are the most used versions in production: simple API, active community, built-in validation, and export to ONNX/TensorRT. For tasks with high accuracy requirements and less critical latency, RT-DETR, a transformer-based architecture without NMS, gives better mAP on COCO at comparable speed to YOLOv8l.

A typical mistake when training a detector: dataset of 8000 images, 3 classes, fine-tune YOLOv8m—F1 0.73 on validation. Look at confusion matrix—one class is almost never detected. Cause: imbalance 1:23. Solution: oversampling rare class, focal loss for objectness, augmentations (Mosaic, MixUp disabled for rare class as they "blur" it). Transfer learning is mandatory: pretrained on COCO weights reduces data requirement by 10 times. Fine-tuning on 500–2000 domain images yields a working model in 1–2 days on a single GPU.

For edge deployment: export to ONNX → TensorRT engine. YOLOv8n in TensorRT FP16 on Jetson AGX Orin gives 150+ FPS at P99 latency < 8 ms—3 times faster than ONNX Runtime without TensorRT. On server A10G: 700+ FPS for YOLOv8n in TensorRT INT8.

How Does Fine-Tuning YOLO Help in Pattern Recognition?

Suppose you need to find micro-defects on a metal surface—a task with high resolution and class imbalance. We use YOLOv8m pretrained on COCO and fine-tune on 2000 proprietary images. Apply augmentations Mosaic, MixUp, random perspective. After 200 epochs, mAP 0.5 reaches 0.93. Key techniques:

• Focal loss for the objectness head—reduces contribution of easily classified examples.
• Class-balanced sampling—equalizes representation of rare classes.
• Test Time Augmentation (TTA)—increases recall by 5–7% through averaging over flips and scales.

Get a consultation on architecture selection for your task—contact us.

Segmentation: SAM, Mask R-CNN, and Instance Segmentation

SAM (Segment Anything Model) from Meta changed the approach to segmentation. SAM 2 works with video, supports object tracking across frames—for interactive object selection by point or bbox, it's the best out-of-the-box choice. For production instance segmentation without interactive prompting, Mask R-CNN or YOLOv8-seg are used. YOLOv8-seg trains like a regular detector with additional masks, convenient in the same pipelines. Semantic segmentation (each pixel is a class) uses SegFormer, DeepLabV3+. SegFormer-B5 provides a good balance of accuracy and speed for satellite imagery or medical segmentation.

Case study: cell segmentation on microscopic images. Dataset of 400 images with manual annotation. Training Mask R-CNN on ResNet-50 backbone gave IoU 0.61—poor. Problem: objects (cells) overlap; standard NMS kills overlapping predictions. Solution: switch to cellpose (specialized architecture for biomedical tasks) + soft-NMS. IoU increased to 0.79.

OCR: When Tesseract Fails

Tesseract is a starting point for simple tasks: printed text, good lighting, straight layout. As soon as there are handwritten elements, non-standard fonts, perspective distortions, or multi-column layouts, Tesseract degrades quickly.

PaddleOCR is a production-grade solution: text block detection + recognition + structural analysis. Works out of the box for 80+ languages, including Russian. Supports tables and complex document structures. TrOCR (Microsoft) is a transformer OCR with strong results on handwritten text. For Russian handwritten text, fine-tuning is needed: the base model is trained mostly on Latin script.

What to Do When Tesseract Cannot Handle Pattern Recognition on Documents?

For tasks like "extract data from invoices/contracts/passports," we use LayoutLMv3 or Donut—these models understand document layout, not just text. Integration via Hugging Face Transformers, fine-tuning on 200–500 annotated documents. Typical pipeline:

1. Preprocessing: deskew, denoising, binarization via OpenCV.
2. Text block detection: PaddleOCR detection or CRAFT.
3. Recognition: PaddleOCR recognition or TrOCR.
4. Post-processing: normalization, validation via regex or LLM for structured fields.

For documents with fixed structure, template matching + OCR by coordinates is often more reliable than an end-to-end solution.

Face Recognition: Identification and Verification

Face recognition = detection + alignment + embedding + matching. Each stage matters.

Detection: RetinaFace or InsightFace for accurate face localization and keypoints. MTCNN is older but reliable. Embedding: ArcFace (InsightFace) is state-of-the-art for face recognition embeddings. Models iresnet50/iresnet100 pretrained on MS1MV3 (5M identities). Embedding vector 512 float32, comparison by cosine similarity. Threshold tuning: decision threshold is a critical parameter. At threshold 0.6, typical FPR on LFW benchmark is 0.001, TPR is 0.985. In production, threshold must be calibrated to the real distribution: people in masks, with changed appearance, different lighting conditions. Liveness detection is mandatory: MiniFASNet—lightweight model on CPU; FaceX-Zoo contains several pretrained liveness detectors.

Video Analytics

Video is a sequence of frames plus a temporal dimension. A naive approach—detecting on every frame—is expensive.

Tracking: ByteTrack and BoT-SORT are the standard for multi-object tracking. They work on top of any detector, adding persistent IDs to objects across frames—enabling object counting, motion tracking, velocity.

Optimization: not every frame needs processing. For static scenes, detect every 5–10 frames, with tracking in between. For event detection (person entering a zone), background subtraction (OpenCV MOG2) serves as a lightweight pre-filter before neural detection. Action recognition: SlowFast, VideoMAE for action classification. Heavy models—for production use ONNX export + TensorRT or offline processing.

How to Measure Pattern Recognition Model Quality in Production?

Quality monitoring is key to MLOps. We track:

• Prediction confidence distribution.
• Share of low-confidence predictions (indicator of OOD data).
• Drift of input images via feature distribution (embeddings from backbone).

A drop in average confidence from 0.87 to 0.71 over a week is an early signal of distribution shift. NVIDIA Triton Inference Server recommends tracking these metrics via Prometheus. Our certified engineers set up monitoring and guarantee SLA for inference quality.

Deployment of CV Models

For online inference, we use Triton Inference Server (NVIDIA)—production standard for serving CV models. Supports TensorRT, ONNX, PyTorch, dynamic batching, multiple instances. REST and gRPC API. We guarantee stable operation under load.

Edge deployment: ONNX Runtime on ARM/x86 CPU. TensorFlow Lite for mobile devices. OpenVINO for Intel CPU/GPU/VPU—gives 2–3× speedup on Intel hardware compared to ONNX Runtime. After deployment, we hand over the model with documentation and train personnel.

What Is Included in the Work

Deadlines and Cost

A prototype detector on existing data takes 1–2 weeks. Production system with optimization for target hardware takes 4–8 weeks. Full cycle including data labeling (1000–5000 images) takes 2–4 months. Cost is calculated individually for each task. Typical savings from implementing a quality control system can be significant per production line.

We have been in the market for over 5 years and completed 60+ computer vision projects. We will evaluate your project end-to-end—request a consultation to get a quote and technical proposal.