Table of Contents

1. Introduction
2. Precision at the Pixel Level: Pixel Error Detection
3. Building a Synthetic Pixel Error Generator

Introduction

Automating tedious tasks in video quality control (QC) allows creative partners to focus on storytelling. A new method detects pixel-level artifacts, reducing the need for manual visual reviews. The goal is to catch technical errors early, enhancing viewer experience.

Precision at the Pixel Level: Pixel Error Detection

Hot pixels, such as bright or stuck pixels, are hard to catch manually. Detecting hot pixels at scale with high accuracy is crucial. By outputting a continuous map of pixel errors at input resolution, the model identifies error centroids for reporting.

Building a Synthetic Pixel Error Generator

Given the rarity of pixel errors, synthetic data generation is key for initial model training. Two types of errors, symmetrical and curvilinear, are simulated. The model is iteratively refined with real-world data to reduce false positives.

---

Please consider the above summary for the provided content.