DSText V2: A Comprehensive Video Text Spotting Dataset for Dense and Small Text

Recently, video text detection, tracking, and recognition in natural scenes are becoming very popular in the computer vision community.However, most existing algorithms and benchmarks focus on common text cases~(\eg normal size, density) and single scenario, while ignoring extreme video text challenges, \ie{} dense and small text in various scenarios. In this paper, we establish a video text reading benchmark, named DSText V2, which focuses on \textbf{D}ense and \textbf{S}mall text reading challenges in the video with various scenarios. Compared with the previous datasets, the proposed dataset mainly include three new challenges: 1) Dense video texts, a new challenge for video text spotters to track and read. 2) High-proportioned small texts, coupled with the blurriness and distortion in the video, will bring further challenges. 3) Various new scenarios, \eg{} `Game', `Sports', etc. The proposed DSText V2 includes 140 video clips from 7 open scenarios, supporting three tasks, \ie{} video text detection (Task 1), video text tracking (Task 2), and end-to-end video text spotting (Task 3). In this article, we describe detailed statistical information of the dataset, tasks, evaluation protocols, and the results summaries. Most importantly, a thorough investigation and analysis targeting three unique challenges derived from our dataset are provided, aiming to provide new insights. Moreover, we hope the benchmark will promise video text research in the community. DSText v2 is built upon DSText v1, which was previously introduced to organize the ICDAR 2023 competition for dense and small video text.

近年来，自然场景下的视频文本检测、跟踪与识别在计算机视觉领域愈发受到广泛关注。然而当前多数算法与基准数据集均聚焦于常规文本场景（如尺寸、密度适中的文本）及单一应用场景，却忽略了极端视频文本挑战——即多样场景下的密集与小型文本问题。为此，本文构建了一款名为DSText V2的视频文本读取基准数据集，其核心聚焦于多样视频场景下的密集（Dense）与小型（Small）文本读取挑战。相较于现有数据集，本次提出的数据集主要涵盖三大全新挑战：1. 密集型视频文本：这是视频文本检测器在跟踪与识别任务中面临的全新挑战；2. 高占比小型文本：结合视频中的模糊与畸变问题，将进一步提升任务难度；3. 多样全新场景：例如游戏、体育赛事等场景。本次构建的DSText V2数据集共包含来自7类开放场景的140段视频片段，支持三大任务：视频文本检测（任务1）、视频文本跟踪（任务2）以及端到端视频文本定位识别（任务3）。本文详细阐述了该数据集的统计信息、相关任务及评估协议，并汇总了实验结果。尤为关键的是，本文针对该数据集衍生出的三大独特挑战展开了全面调研与分析，旨在为领域内研究提供全新视角。此外，本基准数据集期望能够推动计算机视觉社区内的视频文本研究发展。DSText V2基于此前为组织ICDAR 2023密集与小型视频文本竞赛而推出的DSText V1构建而成。

搜集汇总

背景与挑战

背景概述

DSText V2是一个专注于视频中密集和小文本识别的综合数据集，包含140个来自7个不同场景的视频片段，旨在解决现有算法在极端文本挑战中的不足。该数据集支持视频文本检测、跟踪和端到端识别三个任务，是DSText v1的扩展版本，用于促进计算机视觉社区的视频文本研究。