It was taken by Logitech HD Pro C920 webcam. The SSDN dataset contains 13,600 digit images of 10 digits. The 2,000 normal digit images without inherent tilt and lighting conditions were divided into 1,600 images and 400 images for training and testing, respectively.
Moreover, there are four testing image categories– (i) rotate, (ii) scales, (iii) perspective, and (iv) illumination effect. In the rotate category, the image was taken by tilt the camera about −80◦ to 80◦ with 20◦ steps for 5,000 digit images (4,000 for training and 1,000 for testing). The scaled image was taken with a variety of distance between the camera and the seven-segment display from 25 to 150 meters with 25 steps for 4,000 digit images the scales category. The perspective category contained 5,000 digit image data which was skew tilt the seven segment display to left and right with 20◦ to 60◦ by 20◦ steps. For the illumination category, 500 digit images for testing were taken in a dark environment with a diversity of light incidents. The image sizes ranged from 3 × 7 to 36 × 52 pixels and 64 illumination conditions.
It is free to use the SSDN Database for academic researches. All publications which use this database should acknowledge the use of “the SSDN Database” and reference to
- S. Popayorm, T. Titijaroonroj, T. Phoka and W. Massagram, “Data Augmentation Based on Multiscale Radon Transform for Seven Segment Display Recognition,” 2020 12th International Conference on Knowledge and Smart Technology (KST), Pattaya, Chonburi, Thailand, 2020, pp. 47-51.
- S. Popayorm, T. Titijaroonroj, T. Phoka and W. Massagram, “Seven Segment Display Detection and Recognition using Predefined HSV Color Slicing Technique,” 2019 16th International Joint Conference on Computer Science and Software Engineering (JCSSE), Chonburi, Thailand, 2019, pp. 224-229.
All the images can be downloaded below.
Please contact Dr. Taravichet Titijaroonroj (taravichet@it.kmitl.ac.th) if you have any questions or comments about this database.