OmniTrack: Real-time detection and tracking of objects, text and logos in video

TL;DR

OmniTrack is a real-time, robust algorithm that combines deep learning detection with optical flow to track objects, text, and logos in videos efficiently, suitable for standard definition video processing.

Contribution

The paper introduces OmniTrack, a novel real-time detection and tracking algorithm that integrates optimized deep learning and optical flow methods for multiple object types.

Findings

Runs in real-time on standard definition video

Achieves high accuracy in detecting objects, text, and logos

Optimized for performance with asynchronous processing

Abstract

The automatic detection and tracking of general objects (like persons, animals or cars), text and logos in a video is crucial for many video understanding tasks, and usually real-time processing as required. We propose OmniTrack, an efficient and robust algorithm which is able to automatically detect and track objects, text as well as brand logos in real-time. It combines a powerful deep learning based object detector (YoloV3) with high-quality optical flow methods. Based on the reference YoloV3 C++ implementation, we did some important performance optimizations which will be described. The major steps in the training procedure for the combined detector for text and logo will be presented. We will describe then the OmniTrack algorithm, consisting of the phases preprocessing, feature calculation, prediction, matching and update. Several performance optimizations have been implemented there as well, like doing the object detection and optical flow calculation asynchronously. Experiments show that the proposed algorithm runs in real-time for standard definition ($720x576$) video on a PC with a Quadro RTX 5000 GPU.