TextBoxes++: A Single-Shot Oriented Scene Text Detector

By Minghui Liao, Baoguang Shi, Xiang Bai. 2018

800 citations recorded by the source corpus.

Scene text detection is an important step of scene text recognition system and also a challenging problem. Different from general object detections, the main challenges of scene text detection lie on arbitrary orientations, small sizes, and significantly variant aspect ratios of text in natural images. In this paper, we present an end-to-end trainable fast scene text detector, named TextBoxes++, which detects arbitrary-oriented scene text with both high accuracy and efficiency in a single network forward pass. No post-processing other than efficient non-maximum suppression is involved. We have evaluated the proposed TextBoxes++ on four public data sets. In all experiments, TextBoxes++ outperforms competing methods in terms of text localization accuracy and runtime. More specifically, TextBoxes++ achieves an f-measure of 0.817 at 11.6 frames/s for <inline-formula> <tex-math notation="LaTeX">$1024 \times 1024$ </tex-math></inline-formula> ICDAR 2015 incidental text images and an f-measure of 0.5591 at 19.8 frames/s for <inline-formula> <tex-math notation="LaTeX">$768 \times 768$ </tex-math></inline-formula> COCO-Text images. Furthermore, combined with a text recognizer, TextBoxes++ significantly outperforms the state-of-the-art approaches for word spotting and end-to-end text recognition tasks on popular benchmarks. Code is available at: <uri>https://github.com/MhLiao/TextBoxes_plusplus</uri>.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Media Technology

How researchers see this paper

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

In Artificial Intelligence

1. Chunhua Shen
2. Xiang Bai
3. Qi Wu
4. Jianfeng Gao
5. Cong Yao
6. Xue Yang
7. Jiebo Luo
8. Qi Tian

In Computer Vision and Pattern Recognition

1. Xiang Bai
2. Chunhua Shen
3. Luc Van Gool
4. Dacheng Tao
5. Ming-Hsuan Yang
6. Yu Qiao
7. Cong Yao
8. Junchi Yan

In Media Technology

1. Cong Yao
2. Xiang Bai
3. Minghui Liao
4. Xue Yang
5. Xin He
6. Junchi Yan
7. Guisong Xia
8. Shangbang Long

Citations in the corpus

• Scene Text Visual Question Answering (2019, methodology): “Several works have been proposed addressing text detection such as [36, 35, 60, 21] which are mostly based on Fully Convolutional Neural Networks.”
• DOTA: A Large-Scale Dataset for Object Detection in Aerial Images (2017, neutral): “Then R-CNN’s output target T = {(txi, tyi), i = 1, 2, 3, 4} is calculated as, txi = (gxi−xi)/w, tyi = (gyi−yi)/h, where w = xmax−xmin, and h = ymax − ymin, similar as [13].”
• Dynamic Anchor Learning for Arbitrary-Oriented Object Detection (2020, methodology): “To localize rotated objects, the preset rotated anchor and additional angle prediction are adopted in the literature (Liu, Ma, and Chen 2018; Ma et al. 2018; Liao, Shi, and Bai 2018; Liu et al. 2017).”
• Object Detection in Aerial Images: A Large-Scale Benchmark and Challenges (2021, background): “OBB Head To predict OBB, the previous Faster R-CNN OBB [14] and Textboxes++ [62] modified RoI Head of Faster R-CNN and the Anchor Head of the single-shot detector (SSD), respectively, to regress quadrangles.”
• Confidence-aware Non-repetitive Multimodal Transformers for TextCaps (2020, positive): “One way of text detection method is to use box regression adapted from popular object detectors (Liao, Shi, and Bai 2018).”
• Localization Distillation for Object Detection (2022, background): “, object tracking [1, 50], instance segmentation [2, 15] and scenetext detection [32, 51].”
• CORD: A Consolidated Receipt Dataset for Post-OCR Parsing (2019, methodology): “1 Introduction Optical character recognition (OCR) is a technique for converting images of characters into digitized texts [1, 2].”
• Dense Label Encoding for Boundary Discontinuity Free Rotation Detection (2020, methodology): “RRPN [26], TextBoxes++ [18] and RRD [19] and FOTS [23] are some advanced methods for scene text detection.”
• Character Region Awareness for Text Detection (2019, background): “Recently, scene text detectors based on deep learning have shown promising performance [8, 40, 21, 4, 11, 10, 12, 13, 17, 24, 25, 32, 26].”
• Scene Text Detection and Recognition: The Deep Learning Era (2018, methodology): “line: More recent methods follow a 2- step pipeline, consisting of an end-to-end trainable neural network model and a post-processing step that is usually much simpler than previous ones [48]2, [64], [79], [80]3, [93], [131]4, [94]5, [102]6, [78], [83]7, [20], [93], [177]. These methods mainly draw ”
• Learning Modulated Loss for Rotated Object Detection (2019, methodology): “To avoid the inherent regression inconsistency, the eight-parameter representation has been recently developed [42, 46, 47].”
• Rotation-Sensitive Regression for Oriented Scene Text Detection (2018, background): “Thus, we follow a scheme in [25], which determine the first point based on the distances from its corresponding maximum horizontal bounding box.”
• Efficient and Accurate Arbitrary-Shaped Text Detection With Pixel Aggregation Network (2019, background): “TextBoxes++ [26] further regresses quadrangles instead of horizontal bounding boxes for multioriented text detection.”
• Learning RoI Transformer for Oriented Object Detection in Aerial Images (2019, background): “often approached by an oriented and densely packed object detection task [37, 31, 12], which is new while wellgrounded and have attracted much attention in the past decade [27, 30, 26, 18, 1].”