AI-Powered Nutritional Analysis from Food Photos

AI-Powered KBZHU and Meal Composition Analysis from a Photo

Taking a picture of your plate and getting accurate calories, protein, fat, and carbs—this is a challenge dozens of startups are tackling. In practice, accuracy depends on a five-step pipeline: segmentation → classification → volume estimation → database mapping → calculation. The weak link is volume estimation: even state-of-the-art depth estimation models with a reference object produce a 25–35% weight error. Our team has been working on foodtech projects since 2014, delivering 20+ computer vision systems for food recognition. We integrated LiDAR support, cutting the error to 12–18%—twice as good as monocular approaches. Our solutions have already helped clients reduce nutritional counseling costs by 40–60%.

Pipeline: Five Steps from Photo to KBZHU

1. Meal segmentation — isolate individual components on the plate: side, meat, sauce.
2. Classification — identify each component: borscht, chicken breast, buckwheat.
3. Portion volume estimation — the hardest and most narrow part (more details below).
4. Database mapping — USDA FoodData Central, OpenFoodFacts, restaurant TTK.
5. Nutrient calculation — weight × composition per gram.

The weak link is step 3. A classifier can identify borscht with 0.91 accuracy, but if the portion weight is off by ±40%, the final KBZHU will be incorrect.

Why Volume Estimation Is the Hardest Task

Reference Object: Cheap but Inaccurate

The user places a standard-sized card (business card, bank card, QR marker) next to the plate. The system computes scene scale and estimates area. For volume, multiple images (Structure from Motion) are needed, which is unrealistic in a mobile app—users won't walk around the plate.

Depth Estimation: Monocular vs. LiDAR

Monocular models (DPT, ZoeDepth) produce a relative depth map. With a reference object for scaling, RMSE for weight is 25–35%. Acceptable for fitness tracking, unacceptable for dietary control.

LiDAR scanner (iPhone 12 Pro+, iPad Pro) captures real surface coordinates. RMSE for weight is 12–18%—twice as accurate (confirmed by Apple ResearchKit).

Comparison of Volume Estimation Methods

How We Build Accurate Classification Models

Public datasets like Food-101 and UEC Food-256 cover Western dishes, but not Russian cuisine. We use crowdsourcing with nutritionist verification: collect 500–1000 photos per class, label components and weight ratios. This gives 85–88% top-1 accuracy for 80–120 categories.

Accuracy is limited not by CV but by the quality of the food composition database. We combine USDA FoodData Central, OpenFoodFacts, and restaurant TTK cards. For B2B clients, the biggest gain comes from loading their own technological cards—calculations use real recipes.

Which Application Architecture Works for Foodtech?

On-device inference: Core ML (iOS) / TFLite (Android) for segmentation and classification. Latency <300 ms, works offline. Depth estimation is optionally server-side for accuracy. API for integration with dietetic software—fully documented.

What the Work Includes

• Data audit and accuracy requirements definition.
• Model development and fine-tuning (segmentation, classification, depth).
• Integration of composition database (USDA / TTK / custom).
• Mobile app iOS/Android with on-device inference.
• API for integration.
• Documentation and team training.
• Pilot support (1 month).

Timeline

MVP for one cuisine (100–150 dishes), mobile app iOS/Android: 8–12 weeks. Full platform with custom database, LiDAR support, and dietetic software integration: 4–6 months.

Contact us for a preliminary audit—we will assess your project and propose the optimal solution. We guarantee transparency at every stage, and our ISO 13485 certification confirms our experience in precise medical measurements. Request a consultation for your project today.

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.