AI-Powered Restoration of Archival Video

AI Restoration of Archival Video

Old film scanned at 2K, but the result has scratches, flicker, grain, blurred frames, and faded colors. Classic restoration in DaVinci Resolve requires manual treatment of each defect. We have developed a neural network pipeline that automatically handles 80% of this work, leaving final correction to a human. Order archival video restoration — we will assess your project for free. Our experience: over 50 successful projects, years of practice with AI restoration. This article explains how the pipeline works and which models deliver maximum quality gains.

Which Defects Does the Neural Network Address?

Film suffers from many defects, each requiring its own architecture. Grain and noise — film grain is statistically different from digital, and specialized models (Real-ESRGAN, BSRGAN, NAFNet) remove it without losing textures. Scratches and artifacts — video inpainting via ProPainter or E2FGVI automatically detects and fills defects from neighboring frames while maintaining temporal consistency. Low resolution — upscaling with HAT lifts 480p to 4K, yielding PSNR gains of up to 1.8 dB over Real-ESRGAN. Missing frames — interpolation via RIFE or FILM restores smoothness. Faded colors — temporally stable colorization via DDColor with expert oversight.

Why Is the Order of Operations Critical?

Order is critical — wrong sequence multiplies artifacts. If you upscale before removing grain, Real-ESRGAN amplifies grain, mistaking it for details. Here is the proven scheme:

1. Scratch detection and removal (inpainting)
   ↓
2. Denoising / grain removal
   ↓
3. Deflickering (normalize brightness between frames)
   ↓
4. Resolution upscaling
   ↓
5. Frame interpolation (if needed)
   ↓
6. Colorization (if black-and-white original)
   ↓
7. Final temporal smoothing

The most common mistake in "quick" pipelines — skipping deflickering before upscaling leads to flicker artifacts.

Our Case: Documentary Archive

One client — a museum with 60 minutes of documentary material from 16mm film. Heavy grain (equivalent to ISO ~3200), vertical scratches on 15% of frames, intermittent fade-out artifacts. We applied a pipeline with a trained U-Net for scratch detection on synthetic data, ProPainter for inpainting with a temporal window of 10 frames, NAFNet for denoising, HAT-L for upscaling 2K→4K. Colorization was not needed. Processing time on A100 80GB: 4.2 hours for 60 minutes at 1920×1080. Final VMAF rose from 78 to 91, experts confirmed the result. This saved the client 70% of the time compared to manual restoration. We guarantee this approach for your archives.

What Is Included in the Work?

• Analysis of source material: defect assessment, selection of model combination
• Automated processing via a calibrated pipeline, adjusted to content
• Final QA with metrics (VMAF, PSNR, LPIPS) and manual artifact correction
• Pipeline documentation and archiving recommendations
• Possibility of fine-tuning models for your archive's specifics (fine-tuning on your frames)

We guarantee deadlines and quality. Contact us for an assessment — get a free consultation.

Comparison of Approaches

Real-ESRGAN surpasses traditional bicubic methods in texture restoration quality. We use it after grain removal for maximum detail preservation.

How Does AI Restoration Save Time?

Instead of 20–40 hours of manual work (DaVinci Resolve), our pipeline processes an hour of video in 4–8 hours, raising VMAF from 78 to 91. ProPainter performs frame inpainting in seconds — manual retouching of each frame takes hours. For long archives, this reduces costs by 5–10 times. Additionally, batch inference on A100 allows parallel processing of multiple segments.

Limitations

The neural network cannot restore what is not in the frame — if a frame is completely destroyed, inpainting hallucinates content from neighboring frames. Such frames are flagged for manual verification. Colorization of historical material requires expert oversight — the model does not know the actual colors of past objects. It is important to include this stage in the budget.

Processing Timelines

Cost is calculated individually. Our certified engineers guarantee results. Get a consultation — discuss your archive.

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.