AI Object Tracking System for VFX

AI Object Tracking for VFX

We develop AI-based object tracking systems for VFX that solve the problem of track loss under occlusion and motion blur. Classic tools like Mocha Pro and PFTrack require manual point placement and break when blur exceeds 15 pixels. Our neural network solutions (CoTracker, DiffusionTrack, FoundPose) provide stable tracking even with 70% occlusion. We have 5+ years of experience in VFX tracking and over 10 successful projects for commercials and films. Neural tracking is 3x faster than classic methods with similar accuracy, and pricing is calculated individually based on material complexity. Post-production budget savings reach up to 60%.

Tracking Types and Tools

2D point tracking — tracking reference points for stabilization or match move. Modern neural tracking with CoTracker (Meta) tracks 256+ points simultaneously, accounting for mutual dependencies. It uses a transformer with temporal attention, providing robustness under partial occlusion. — Meta Research

Planar tracking — tracking a plane (wall, car side, screen) for graphic insertion. Neural version: DiffusionTrack, or a hybrid of homography + deep features from SuperGlue for matching.

6DoF object tracking — tracking 3D position and orientation of an object. FoundPose, FoundTrack work with a CAD model or learned prior.

Human body tracking — MediaPipe Holistic (33 skeleton points + hands + face), OpenPose, SMPL-X for full 3D body reconstruction from monocular video. Used for markerless motion capture.

Why Neural Tracking Is More Robust Than Classic?

Classic feature-based tracking (Lucas-Kanade, KLT) loses points at motion blur >15 px, occlusion >40% of the object, or sudden lighting changes. Neural approaches win through learned descriptors and visibility prediction. For example, CoTracker maintains tracking up to 70% occlusion — the model predicts whether a point is visible in the current frame and doesn't attempt to search where it isn't. When the object re-emerges from occlusion, tracking resumes via re-identification.

import torch
from cotracker.predictor import CoTrackerPredictor

model = CoTrackerPredictor(checkpoint='cotracker2.pth')
model = model.cuda()

# video: (1, T, 3, H, W) tensor
# queries: (1, N, 3) - (t, x, y) for each point
tracks, visibility = model(video, queries=queries)
# tracks: (1, T, N, 2) - coordinates of N points in all T frames
# visibility: (1, T, N) - visibility probability

How to Integrate Tracking into a Post-Production Pipeline?

We output tracking in formats compatible with industry software: .nuke scripts or .abc (Alembic) files with animation curves for Nuke, After Effects, Blender. For match move, we use .chan format with camera transform data. Automation: Python API for Nuke allows direct writing of tracking data into Tracker4 nodes without manual transfer. This speeds up integration and reduces error risks when handing off between departments.

Step-by-step:

1. Prepare the video file and define target objects.
2. Choose tracking type (2D, planar, 6DoF) based on the task.
3. Run the neural model on GPU — average processing time for 5 minutes of video is ~2 hours.
4. Verify track quality via visualization.
5. Export data in required format.
6. Integrate into the scene via plugin or script.

What's Included

• Material analysis and approach selection (2D/planar/6DoF/human)
• Neural model development and tuning for the specific scene
• Tracking execution with intermediate iterations
• Data export in required formats (Nuke, AE, Blender)
• Documentation of used models and configuration
• Training your team on the delivered scripts
• Technical support during post-production

Timelines

Pricing is calculated individually based on material complexity. We guarantee track deviation under 1 px on verified frames. Contact us for a project evaluation — we'll select the optimal turnkey solution. Request a consultation to discuss details.

From Our Practice

A 45-second commercial with a camera moving alongside a car on a highway. A logo needed to be "stuck" to the door, accounting for reflections and body deformation due to vibration. Challenge: speed 120 km/h, motion blur at 1/60s shutter, periodic sun flares. Solution: planar tracking via homography with SuperGlue-matching (robust to blur thanks to learned descriptors) + mesh-based deformation (door divided into an 8×4 grid, each node tracked independently) + lighting estimation via EfficientLit. Result: stable tracking over 1080 frames with deviation <0.8 px, 6 hours of automated processing instead of 3 days of manual tracking in Mocha.

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.