TextSnake: A Flexible Representation for Detecting Text of Arbitrary Shapes

Abstract

Driven by deep neural networks and large scale datasets, scene text detection methods have progressed substantially over the past years, continuously refreshing the performance records on various standard benchmarks. However, limited by the representations (axis-aligned rectangles, rotated rectangles or quadrangles) adopted to describe text, existing methods may fall short when dealing with much more free-form text instances, such as curved text, which are actually very common in real-world scenarios. To tackle this problem, we propose a more flexible representation for scene text, termed as TextSnake, which is able to effectively represent text instances in horizontal, oriented and curved forms. In TextSnake, a text instance is described as a sequence of ordered, overlapping disks centered at symmetric axes, each of which is associated with potentially variable radius and orientation. Such geometry attributes are estimated via a Fully Convolutional Network (FCN) model. In experiments, the text detector based on TextSnake achieves state-of-the-art or comparable performance on Total-Text and SCUT-CTW1500, the two newly published benchmarks with special emphasis on curved text in natural images, as well as the widely-used datasets IC-DAR 2015 and MSRA-TD500. Specifically, TextSnake outperforms the baseline on Total-Text by more than 40% in F-measure.

Mask TextSpotter: An End-to-End Trainable Neural Network for Spotting Text with Arbitrary Shapes

“…"Det eval" 3. https://github.com/cs-chan/Total-Text-Dataset/tree/master/ Evaluation Protocol adopts two stricter thresholds, where "tr" and "tp" are set to 0.7 and 0.6 respectively. Note that although [48] reported detection results on Total-Text, it used a different evaluation protocol. Thus, for a fair comparison, we do not list their performance in Table 5, and just report the results here for reference (precision: 82.7%, recall: 74.5% f-measure: 78.4%).…”

Curved Text Detection in Natural Scene Images with Semi- and Weakly-Supervised Learning

“…In [6], DeconvNet [15] is used to perform curved text detection. Textsnake [8] models scene text as a sequence of ordered, overlapping disks centered at symmetric axes, each of which is associated with potential variable radius and orientations. This method performs well on arbitrary shape text but is slowed down by a complex post processing.…”

“…The other three settings are naive strategy, filter strategy and local strategy which are described in section III-C. F-measure versus training round for different strategies on CTW1500 test set. SOTA here corresponds to [8] . Figure 4.…”

“…Recently, various methods [8][9] [6][10] based on semantic segmentation [11] and instance segmentation [12] are proposed to detect text of arbitrary shape. They demonstrate superior performance and generalization on various datasets.…”

A New Perspective for Flexible Feature Gathering in Scene Text Recognition Via Character Anchor Pooling

“…There are various applications, including instant translation, robot navigation, industrial automation, and traffic sign reading for autonomous vehicles. More recently, the detection and recognition of irregular text, e.g., text arranged in a curved line, has attracted much attention [2,3].…”