EAST: An Efficient and Accurate Scene Text Detector

Abstract

Previous approaches for scene text detection have already achieved promising performances across various benchmarks. However, they usually fall short when dealing with challenging scenarios, even when equipped with deep neural network models, because the overall performance is determined by the interplay of multiple stages and components in the pipelines. In this work, we propose a simple yet powerful pipeline that yields fast and accurate text detection in natural scenes. The pipeline directly predicts words or text lines of arbitrary orientations and quadrilateral shapes in full images, eliminating unnecessary intermediate steps (e.g., candidate aggregation and word partitioning), with a single neural network. The simplicity of our pipeline allows concentrating efforts on designing loss functions and neural network architecture. Experiments on standard datasets including ICDAR 2015, COCO-Text and MSRA-TD500 demonstrate that the proposed algorithm significantly outperforms state-of-the-art methods in terms of both accuracy and efficiency. On the ICDAR 2015 dataset, the proposed algorithm achieves an F-score of 0.7820 at 13.2fps at 720p resolution.

Efficient Scene Text Detection with Textual Attention Tower

“…The detection branches in our proposed method are denoted as DET. Inspired by [1,9], we use rotated box (RBOX) to describe text regions. Thus the DET branch is simply 1 × 1 convolutions to map final feature to detections.…”

“…Moreover, bounding a single text object with an up-right detection box may lead to a low IoU and detection quality. Several approaches [1,2,3] have already presented impressive successes on various public benchmarks and competitions.…”

“…The key of text detection is designing features to distinguish text from backgrounds. Recently, Convolutional Neural Networks (CNN) based methods such as EAST [1] and IncepText [4] have achieved the state-of-the-art performance for text detection. Like other computer vision tasks, deeper networks provide better performances.…”

Chinese text-line detection from web videos with fully convolutional networks

“…The architecture of the proposed network in our paper is inspired by a deconvolution network for Semantic Segmentation [16] and EAST [17]. Deconvolution network plays an important role in image semantic segmentation.…”

“…There are several strategies which can be referred to solve this problem, such as Online Hard Negative Mining [21] and Weighted Softmax Loss. In this work, we use the class-balanced cross-entropy loss function which is effective for several systems [17,22,23]. The function is defined with…”