FCENet: An Instance Segmentation Model for Extracting Figures and Captions From Material Documents

Liu, Yingli; Si, Changkai; Jin, Kai; Shen, Tao; Hu, Meng

doi:10.1109/access.2020.3046496

Cited by 11 publications

(5 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The segmentation-based detection method regards the detection problem as a text classification problem by predicting text instances at the pixel level and then detecting multiangle, irregular text using the concept of semantic segmentation. Mainstream network topologies primarily consist of PSENet [30], DB [31], and FCENet [32], which are superior to regression-based detection methods. Other academics have proposed combined regression and segmentation detection methods, which would make it easier to find irregular text, reduce false alarms, and make text more robust at different scales, among other things.…”

Section: Text Detectionmentioning

confidence: 99%

MTR-SAM: Visual Multimodal Text Recognition and Sentiment Analysis in Public Opinion Analysis on the Internet

et al. 2023

View full text Add to dashboard Cite

Existing methods for monitoring internet public opinion rely primarily on regular crawling of textual information on web pages but cannot quickly and accurately acquire and identify textual information in images and videos and discriminate sentiment. The problems make this a challenging research point for multimodal information detection in an internet public opinion scenario. In this paper, we look at how to dynamically monitor the internet opinion information (mostly images and videos) that different websites post. Based on the most recent advancements in text recognition, this paper proposes a new method of visual multimodal text recognition and sentiment analysis (MTR-SAM) for internet public opinion analysis scenarios. In the detection module, a LK-PAN network with large sensory fields is proposed to enhance the CML distillation strategy, and an RSE-FPN with a residual attention mechanism is used to improve feature map representation. Second, it proposes that the original CTC decoder be replaced with a GTC method to solve earlier problems with text detection at arbitrary rotation angles. Additionally, the performance of scene text detection for arbitrary rotation angles is improved using a sinusoidal loss function for rotation recognition. Finally, the improved sentiment analysis model is used to predict the sentiment polarity of the text recognition results. The experimental results show that the new method proposed in this paper improves recognition speed by 31.77%, recognition accuracy by 10.78% on the video dataset, and the F1 score of the multimodal sentiment analysis model by 4.42% on the self-built internet public opinion dataset (lab dataset). The method proposed provides significant technical support for internet public opinion analysis in multimodal domains.

show abstract

Section: Text Detectionmentioning

confidence: 99%

MTR-SAM: Visual Multimodal Text Recognition and Sentiment Analysis in Public Opinion Analysis on the Internet

et al. 2023

View full text Add to dashboard Cite

show abstract

“…Group 1 can be divided into 2 subgroups: (1) Subgroup 1 is documents created on plain paper without an overlaid pattern. There are relevant research topics such as text/non-text classification in online handwritten notes [16], the 2D chemical structures recognition in document images [17], detecting math www.ijacsa.thesai.org equations in scientific document images [18], the Arabic word recognition of historical documents images [19], the Vietnamese character recognition for verifying ID card [20], document zoning for document layout analysis [21], analysis of the structure of the musical document image [22], bibliographic reference extraction [23], extracting text and figure from document images [24,25], document localization in natural scene images [26], and table detection and segmentation in document images [27]. (2) Subgroup 2 is created on plain paper and overlapped patterns.…”

Section: Related Workmentioning

confidence: 99%

The Method of Braille Embossed Dots Segmentation for Braille Document Images Produced on Reusable Paper

Tiendee¹,

Lerdsudwichai²,

Thainimit³

et al. 2022

IJACSA

View full text Add to dashboard Cite

Braille is the language of communication for blind and visually impaired people. Braille characters are embossed at points to convey the meaning. Typically, Braille documents can be produced on plain paper. Braille documents can be created on reusable paper, also known as a third-page paper; this reduces the paper cost, allowing more available documents to stimulate learning for blind or visually impaired persons. This research presents a method of Braille embossed dots segmentation for Braille document images produced on reusable paper to support the availability of cheaper learning material. Initially, Braille documents were imported with a calibrated scanner, Braille document image layer separation was then performed. Followed by edge removal, Braille embossed dot recovery, noise removal, and specify the embossed Braille point. This research was conducted by using four scanners, which scanned Braille documents images under four different lighting conditions. For each lighting condition, the Braille document image area was cropped to the desired size, considering the possible event conditions. They were used to create over 200,000 Braille cells, with over 12 billion patterns. When calculating the average performance under all lighting conditions, the values were Precision 1.0000, Recall 0.7817, Accuracy 0.8545, and F-Measure 0.8756. By effectively using Braille embossed dots segmentation, the process of Braille document recognition will also be efficient.

show abstract

“…This method introduces a progressive scaling algorithm capable of successfully identifying adjacent text instances. In contrast, FCENet [30] directly predicts the Fourier feature vector for each pixel location. Text contours are subsequently reconstructed through the Inverse Fourier Transform (IFT) applied to the Fourier feature vectors of areas surpassing the score threshold.…”

Section: Related Workmentioning

confidence: 99%

Fenet: Feature Enhancement Network for Arbitrary Direction Text Detection

Wang,

Liu,

Han

et al. 2022

SSRN Journal

View full text Add to dashboard Cite

FCENet: An Instance Segmentation Model for Extracting Figures and Captions From Material Documents

Cited by 11 publications

References 31 publications

MTR-SAM: Visual Multimodal Text Recognition and Sentiment Analysis in Public Opinion Analysis on the Internet

MTR-SAM: Visual Multimodal Text Recognition and Sentiment Analysis in Public Opinion Analysis on the Internet

The Method of Braille Embossed Dots Segmentation for Braille Document Images Produced on Reusable Paper

Fenet: Feature Enhancement Network for Arbitrary Direction Text Detection

Contact Info

Product

Resources

About