Research Areas
My primary research interests include the following areas:
- Image Processing and Pattern Recognition
- Document Image Analysis
- Machine Learning
- Machine-based Translation
Downloads
- BCBiD Dataset
BESUS Dataset: Since there is no standard benchmark Bangla handwritten database for writer identification, we have created our own Bangla handwritten database (BESUS Database). The BESUS database consists of images of writings of 55 writers. Every writer has four samples on two different topics. Here the topic for the writings is so chosen that most of the basic Bangla characters are present in the writings. Few samples are shown below. Fifty percent of the total writers of the database are female and the remaining writers are male with the age group varying from 21 years to 23 years.
Related Works:
- Samit Biswas, Amit Kumar Das, "Writer Identification of Bangla handwritings by Radon Transform Projection Profile", DAS-2012, Gold Coast, Queensland, Australia, pp: 215-219, March 2012.
- Samit Biswas, Amit Kumar Das, "Content Independent Writer Identification using Occurrences of Writing Styles for Bangla Handwritings", NCVPRIPG-2011, IEEE CS Press, pp.154-157.
LMIDb: Map document processing research suffers due to non-availability of benchmark public dataset. To help the researchers we have created a dataset named the Land Map Image Database (LMIDb). It consists of a variety of land maps images (446 images at present; scanned in 200 and 300 dpi in TIFF; with 128 English maps, 100 Bangla maps, and 218 Hindi maps) and have been created specifically for the use of Map document image analysis. The texts in these maps are in English and in a number of Indic scripts like Bangla, Hindi etc. It can be used to evaluate the performance of automatic offline map image component (i.e., texts for place names, boundary etc.) extraction methods. For every map image corresponding binarized ground-truth, text only ground-truth and thematic ground truth (presently containing river only, road only and icons only ground truth) image are available in the database. Also, the XML files contains the bounding coordinates of each text block present in the text only groundtruth.
Related Works:
- Sayan Mandal; Samit Biswas and Amit Kumar Das, "Land Map Image Dataset: Ground-Truth And Classification Using Visual And Textural Features", Image Processing & Communications, 19(4), 37-55, 2014. doi: https://doi.org/10.1515/ipc-2015-0024
- Sayan Mandal; Samit Biswas; Amit Kumar Das and Bhabatosh Chanda; "Map image binarization and stitching using extraction of regions", Journal of Theoretical and Applied Computer Science, vol. 9, no. 1, pp. 28-40, 2015.
- Samit Biswas, Amit Kumar Das, and Bhabatosh Chanda. Text Segmentation from Bangla Land Map Images, Image Processing & Communications, 19(1), 21-34, 2014. doi: https://doi.org/10.1515/ipc-2015-0003