13 datasets found

The MSRA-TD500 dataset is a benchmark for scene text detection, containing 700 training images and 200 test images, with multi-lingual, arbitrary-oriented and long text lines.

The proposed scene text image synthesis technique starts with two types of inputs including “Background Images” and “Source Texts” as illustrated in column 1 and 2 in Fig. 1.

The ICDAR-2017 Robust Reading Competition dataset contains images with text in various fonts, sizes, and orientations.

The ICDAR-Art dataset contains a total of 10,166 images, 5603 images in the training set and 4563 images in the testing set.

The CTW1500 dataset contains 1,500 images, of which 1,000 are for training and 500 are for testing. Each image has at least one curved text.

The MSRA-TD500 dataset contains 500 natural scene images, of which 300 are for training and 200 are for testing.

The Total-Text dataset contains the text of various shapes, including horizontal, multi-orientational, and curved.

Text detection in natural images

The COCO-Text dataset contains text in images, with a total of 120,000 images and 1,000 text annotations.

ICDAR2015 dataset consists of 1,670 images (17,548 annotated text regions) acquired using the Google Glass.

ICDAR2013 dataset is obtained from the Robust Reading Challenges 2013.

Total-Text is a dataset for word-level arbitrary-shaped English text detection, containing 1,255 images for training and 300 images for testing.

The dataset contains images of text in natural scenes, including street signs, logos, and product labels.