African foods for deep learning-based food recognition systems dataset

Area of Science:

• Computer Science
• Human-Computer Interaction (HCI)
• Cultural Studies

Background:

• African foods possess significant socio-cultural importance, often transcending borders through migration and social integration.
• The Human-Computer Interaction (HCI) community has historically shown limited representation of African food research.
• Cultural diversity within Africa is vast, encompassing language, cuisine, traditions, and beliefs.

Purpose of the Study:

• To address the underrepresentation of African food in HCI research by contributing novel datasets.
• To promote cultural inclusion within the HCI community by highlighting African culinary heritage.
• To provide researchers with valuable visual data for developing culturally relevant HCI applications.

Main Methods:

• Collected image datasets of popular Cameroonian foods (Ekwang, Eru, Ndole) and Ghanaian foods (Jollof Rice, Palm-nut Soup, Waakye).
• Sourced data from diverse platforms including YouTube, Facebook, field research (restaurants), and Creative Commons image repositories.
• Validated the universal recognition of selected dishes within their respective countries through local consultations.

Main Results:

• Compiled comprehensive datasets with a significant number of images: Ekwang (204), Eru (206), Ndole (205), Jollof Rice (347), Palm-nut Soup (392), and Waakye (400).
• Provided detailed meta-data descriptions and quality assessments for the collected food image datasets.
• Identified and outlined future research opportunities for the HCI community utilizing these datasets.

Conclusions:

• The curated African food datasets offer a valuable resource for the HCI community.
• This contribution aims to foster greater cultural diversity and inclusion in HCI research and development.
• The datasets can support the creation of more culturally sensitive and relevant technological solutions for diverse global users.