What does scene text tell us?

Uchida, Seiichi; Shinahara, Yuto

doi:10.1109/icpr.2016.7900267

Cited by 2 publications

(2 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Those methods will be useful for analyzing the semantics of scene texts. In [20], the words from scenery images were fed to word2vec and then their semantic vectors are classified into several clusters by k-means, in order to understand what kind of information is conveyed by scene texts. It is important to note that we can apply the word embedding methods not only to the words in scene texts but also to the object class names (as we see in Section IV-D).…”

Section: Techniques Required For Our Experimental Surveymentioning

confidence: 99%

“…Table I shows the 40 most frequent words, as one of the most primitive statistics of the words captured in the real images; specifically, this is the list of the words with the 40 largest i g i,k . As suggested in [20], the words captured in scenery images will have a different trend from the words in messages like Wikipedia and British National Corpus (BNC) dataset 7 . The parenthesized number in Table I is the frequency rank of the word in BNC (after the same post-processing steps, such as the stop-word removal).…”

Section: A Frequent Wordsmentioning

confidence: 99%

See 1 more Smart Citation

Label or Message: A Large-Scale Experimental Survey of Texts and Objects Co-Occurrence

Takeshita

Shioyama

Uchida

2020

Preprint

Self Cite

View full text Add to dashboard Cite

Our daily life is surrounded by textual information. Nowadays, the automatic collection of textual information becomes possible owing to the drastic improvement of scene text detectors and recognizer. The purpose of this paper is to conduct a large-scale survey of co-occurrence between visual objects (such as book and car) and scene texts with a large image dataset and a state-of-the-art scene text detector and recognizer. Especially, we focus on the function of "label" texts, which are attached to objects for detailing the objects. By analyzing co-occurrence between objects and scene texts, it is possible to observe the statistics about the label texts and understand how the scene texts will be useful for recognizing the objects and vice versa.

show abstract

Section: Techniques Required For Our Experimental Surveymentioning

confidence: 99%

Section: A Frequent Wordsmentioning

confidence: 99%