Coffee and cashew nut dataset: A dataset for detection, classification, and yield estimation for machine learning...

Area of Science:

• Agricultural Science
• Computer Vision
• Machine Learning

Background:

• Traditional crop yield estimation methods are expensive, inefficient, and inaccurate.
• Accurate crop yield prediction is vital for farm management and market planning.
• A significant data gap exists for machine learning-based crop yield estimation, especially in sub-Saharan Africa.

Purpose of the Study:

• To address the lack of specialized datasets for machine learning in agriculture.
• To provide curated image datasets for coffee and cashew nut crops.
• To facilitate the development of automated crop yield estimation techniques.

Main Methods:

• Collected high-resolution aerial imagery using an Unmanned Aerial Vehicle (UAV) over nine months.
• Acquired datasets during two crop harvest seasons in Uganda.
• Annotated images with object classes relevant to crop health and yield (e.g., fruit maturity, tree health).

Main Results:

• Curated datasets comprising 3000 coffee and 3086 cashew nut images (6086 total).
• Coffee dataset includes annotations for unripe, ripening, ripe, spoilt, and coffee_tree.
• Cashew nut dataset includes annotations for tree, flower, premature, unripe, ripe, and spoilt.

Conclusions:

• The presented datasets are valuable resources for machine learning applications in agriculture.
• These datasets can support tasks such as yield estimation, disease diagnosis, and maturity analysis.
• The data collection in Uganda specifically targets the under-resourced sub-Saharan African region.