Open Datasets and Tools for Arabic Text Detection and Recognition in News Video Frames

Abstract

Abstract:Recognizing texts in video is more complex than in other environments such as scanned documents. Video texts appear in various colors, unknown fonts and sizes, often affected by compression artifacts and low quality. In contrast to Latin texts, there are no publicly available datasets which cover all aspects of the Arabic Video OCR domain. This paper describes a new well-defined and annotated Arabic-Text-in-Video dataset called AcTiV 2.0. The dataset is dedicated especially to building and evaluating Arabic video text detection and recognition systems. AcTiV 2.0 contains 189 video clips serving as a raw material for creating 4063 key frames for the detection task and 10,415 cropped text images for the recognition task. AcTiV 2.0 is also distributed with its annotation and evaluation tools that are made open-source for standardization and validation purposes. This paper also reports on the evaluation of several systems tested under the proposed detection and recognition protocols.

A Survey of OCR in Arabic Language: Applications, Techniques, and Challenges

“…Type of Content Availability Size of Dataset ACTIV2 [12] Embedded words Public 10,415 text images QTID [13] Synthetic words Private 309,720 words and 249,428 characters IFN/ENIT [14] Handwritten words Public 115,000 words and 212,000 characters AHDB [15] Handwritten words and digits Private 30,000 words APTI [16] Printed words Public 113,284 words and 648,280 characters HACDB [17] Handwritten characters Public 6600 characters and 50 writers UPTI [18] Printed text lines Public 10,000 text lines Digital Jawi [19] Jawi paleography images Public 168 words and 1524 characters KHATT [20] Handwritten text lines Public 9327 lines, 165,890 words and 589,924 characters ALIF [21] Embedded text lines Upon request 1804 words and 89,819 characters ACTIV [22] Embedded text lines Public 4824 lines and 21,520 words SmartATID [23] Printed and handwritten pages Public 9088 pages Degraded historical [24] Handwritten documents Public 10 handwritten images and 10 printed images Printed PAW [25] Printed subwords Upon request 415,280 unique words and 550,000 sub words Checks [26] Handwritten subwords and digits Private 29,498 subwords and 15,148 digits Numeral [27] Handwritten digits Public 21,120 digits and 44 writers Forms [28] Handwritten characters Private 15,800 characters and 500 writers KAFD [29] Printed pages and lines Public 28,767 pages and 644,006 lines AHDBIFTR [30] Handwritten images Public 497 word images and 5 writers ARABASE [31] Handwritten text Public 47,000 words and 500 free Arabic sentences CEDAR [32] Handwritten pages Private 20,000 words, 10 writers, and 100 documents CENPARMI [26] Handwritten subwords and digits Public 6000 digit images Shafi and Zia [33] surveyed automatic Urdu text recognition techniques and described the algorithms, techniques, datasets, challenges, and future directions for Urdu OCR. Additionally, [34] reviewed the availability of datasets and suggested more training data to address the unique challenges of OCR systems.…”

A Review of Arabic Text Recognition Dataset

“…The ACTIV 2.0 Dataset (Zayene et al 2018a) is a public dataset that was extracted from 189 video clips, and produces 4,063 key-frames for detection and 10,415 cropped text images for recognition. This dataset is distributed with open-source tools for annotation and evaluation.…”

Arabic Cursive Text Recognition from Natural Scene Images

“…Various efforts have been reported for capturing and preparing the datasets for Arabic text in natural images in the recent past. Some articles presented a survey on available open access datasets and tools specifically designed for Arabic text detection and recognition in video frames captured by news channels [76]. The benchmark dataset for Arabic scene text still requires more effort to standardized the research as far as Arabic scene text analysis is concerned.…”