Description

The proposed food recognition model is currently trained on 20 popular Pakistani dishes, covering a diverse mix of curries, rice, snacks, and grilled items. The dataset includes traditional meals such as aloo keema, chicken biryani, chicken karahi, chicken korma, daal chana, and daal mash, as well as regional specialties like chapli kabab, seekh kabab, shami kabab, and chicken sajji. It also incorporates common fast foods and everyday items, including burger, samosa, fry egg, and fry fish. Rice-based dishes such as beef pulao and protein-rich options like bean curry and chicken roast (leg and chest pieces) further enrich the dataset. Vegetables like bhindi (okra) are also represented. This variety ensures that the model learns to classify foods with high intra-class variability (different cooking styles, plating, and regions) and inter-class similarity (dishes with overlapping visual features), making it robust for real-world food recognition applications in the Pakistani context.