Recognizing Text with Perspective Distortion in Natural Scenes

By Trung Quy Phan, Palaiahnakote Shivakumara, Shangxuan Tian, Chew Lim Tan. 2013

475 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

Citations in the corpus

GIT: A Generative Image-to-text Transformer for Vision and Language (2022, positive): “, 2011), Street View Text-Perspective (SVTP) (Phan et al., 2013), and CUTE80 (CUTE) (Risnumawan et al.”
Structured Multimodal Attentions for TextVQA (2020, methodology): “The robust transformer based network is trained on following datasets: IIIT 5K-Words [29], Street View Text [43], ICDAR 2013 [22], ICDAR 2015 [21], Street View Text Perspective [33], CUTE80 [36] and ArT [10].”
Text Detection and Recognition in Imagery: A Survey (2015, unknown): “[191], Phan et al.”
OCR-Free Document Understanding Transformer (2021, background): “Both are trained with a large-scale datasets including the synthetic images [26,13] and real images [28,47].”
Towards Accurate Scene Text Recognition With Semantic Reasoning Networks (2020, background): “Street View Text-Perspective (SVTP) [27] is also cropped form Google Street View.”
TextOCR: Towards large-scale end-to-end reasoning for arbitrary-shaped scene text (2021, background): “Prior to TextOCR, many OCR datasets exist [38, 59, 35, 24, 23, 45, 48, 8, 61, 34, 57, 41, 50, 40, 9, 54] that propelled the field’s development, but many of these are either relatively small, or focus mostly on outdoor or store-front scenes.”
RobustScanner: Dynamically Enhancing Positional Clues for Robust Text Recognition (2020, methodology): “arbitrary angles. Hence most word patches in this dataset are irregular (oriented, perspective or curved). It contains 4468 patches for training and 2077 for test. Street View Text Perspective (SVTP) [36] consists of 639 word patches, which are cropped from side view snapshots in Google Street View ”
MaskOCR: Text Recognition with Masked Encoder-Decoder Pretraining (2022, methodology): “We evaluate our model on six public scene text datasets: ICDAR 2013 (IC13) [26], Street View Text (SVT) [52], IIIT5K-Words (IIIT5K) [38]), ICDAR 2015 (IC15) [25], Street View Text-Perspective (SVTP) [39], and CUTE80 (CUTE) [43]).”
Mask TextSpotter: An End-to-End Trainable Neural Network for Spotting Text with Arbitrary Shapes (2018, methodology): “are provided by the Task 4.3 of the ICDAR 2015 competition [36]. The images are taken by Google glasses incidentally. There is a large portion of oriented text in this dataset. SVT-Perspective (SVTP) [59] is similar to SVT. However, most of them are distorted by perspective transformation. The test ”
Self-Supervised Learning from Web Data for Multimodal Retrieval (2019, background): “…while synthetic data has proven to be a valuable source of training data for many applications such as pedestrian detection [5], image semantic segmentation [6] and scene text detection and recognition [7, 8], nowadays it is still not easy to generate realistic complex images for some tasks.”
Visual Semantics Allow for Textual Reasoning Better in Scene Text Recognition (2021, unknown): “We test the trained model on six benchmarks including three regular scene-text datasets, i.e. , IC-DAR2013 (Karatzas et al. 2013), IIIT5K (Mishra, Alahari, and Jawahar 2012), SVT (Wang, Babenko, and Belongie 2011), and three irregular text datasets, i.e. , ICDAR2015 (Karatzas et al. 2015), SVTP (Pha”
What Is Wrong With Scene Text Recognition Model Comparisons? Dataset and Model Analysis (2019, methodology): “• SVT Perspective (SP) [21] is collected from Google Street View and contains 645 images for evaluation.”
Show, Attend and Read: A Simple and Strong Baseline for Irregular Text Recognition (2018, background): “Early approaches mainly work in a bottom-up fashion (Wang, Babenko, and Belongie 2011; Mishra, Ala-hari, and Jawahar 2012b; Phan et al. 2013; Yao et al. 2014), in which individual characters are detected ﬁrstly via sliding window or connected components, and then integrated into a word by dynamic…”
Robust Scene Text Recognition with Automatic Rectification (2016, methodology): “The first five rows are taken from SVT-Perspective [29], the rest rows are taken from CUTE80 [30].”
OCRBench: on the hidden mystery of OCR in large multimodal models (2023, unknown): “…text recognition datasets, including (1) Regular Text Recognition: IIIT5K [37], SVT [38], IC13 [39]; (2)Irregular Text Recognition: IC15 [40], SVTP [41], CT80 [42], COCOText(COCO) [43], SCUT-CTW1500 (CTW) [44], Total-Text (TT) [45]; (3) Occlusion Scene Text [46], encompassing weakly occluded scene…”
ASTER: An Attentional Scene Text Recognizer with Flexible Rectification (2019, background): “SVT-Perspective (SVTP) is proposed in [49] for evaluating the performance of recognizing perspective text.”
Scene text detection and recognition: recent advances and future trends (2015, background): “To tackle these challenges, a rich body of approaches have been proposed and substantial progresses have been achieved in recent years [8–20].”