Dataset of 7632 images of 53 Devanagari Alphabet Images Across 144 Spatial Positions in 5×5 Grids with 2×2 Sub grid Localization for Grid based Graphical Password Authentication.

The dataset is a structured image dataset designed to facilitate spatial localization, pattern recognition, and character classification research. It contains high-resolution images of 53 distinct alphabet characters, each systematically placed within a standardized 5×5 grid layout.

Each 5×5 grid consists of 25 individual cells. Within each grid, we define 16 overlapping 2×2 sub-grids. These sub-grids serve as local regions of interest for fine-grained spatial analysis. In each 2×2 sub-grid, there are 9 specific positional locations where an alphabet image can be placed—centered within or slightly offset relative to the sub-grid to provide a range of spatial variation. This results in a total of 144 unique placement positions for each character across the entire 5×5 grid.

The dataset includes an image placed in each of these 144 locations for every alphabet character, leading to a comprehensive total of 7,632 labeled samples (53 characters × 144 positions). All samples are consistent in size and format, and the position of each character is precisely annotated to facilitate supervised learning tasks.

The Devanagari 53 Alphabet dataset is ideal for training and evaluating models on tasks such as character localization, grid-based graphical password , and few-shot learning under positional variation. The structured spatial layout and extensive position coverage also make it suitable for research in visual attention models, object detection benchmarks, and spatially-aware neural architectures.