A Comprehensive Dataset for Bangladeshi Sweets Classification Using Deep Learning

The dataset contains original high-quality JPG photographs of traditional Bangladeshi desserts, organised into 63 class-labelled folders. Images were collected in two stages: (i) traditional mithai (sweets) from local sweet shops in Dhaka, and (ii) pithas collected from the Nabanna festival — a traditional Bengali harvest celebration in late November/early December where handmade pithas using new-crop rice, date-palm jaggery, and seasonal ingredients are prepared. There are no duplicate, low- quality, or superfluous photos; all images are accurately labelled. This extended report expands the scope to 63 categories by documenting additional culturally significant Bangladeshi sweets and pithas not covered in the original dataset. These additional categories include regional specialities and variant preparations that reflect the full breadth of Bangladesh's sweet-making traditions.