Banana Bunch Dataset: Multi-Field Acquisition with Various Environmental Conditions

Dataset overview: This dataset includes images of banana bunches collected from various fields on the island of Madeira, Portugal. The dataset includes four different fields through 2022: Câmara de Lobos on July 25th, Lugar de Baixo on May 17th and 31st, Ponto do Sol on May 20th, June 3rd, August 22nd and October 26th, Santo António on June 7th. The dataset includes a total of 2,179 samples collected under different environmental and lightening conditions. Labels for banana bunches were manually applied to verify how accurate the object detection process was.

Data collection process: The dataset was gathered in 2022 using a variety of smartphone cameras, which includes Samsung Galaxy A12, Galaxy Note 9, One Plus 9 and Iphone 12. The wide range of cameras used to gather images in various environments has improved the resilience and generalisability of possible machine learning models.

Labels and captions: The bunch of bananas visible in the final images underwent a laborious human labelling process in order to receive the correct caption. The generated label collection is used to compare the images, which could offer crucial baseline information for developing and testing object detection models.

Data usage: The goal of this dataset is to make it easier to develop and test machine learning models for detecting banana bunches. This dataset can be used by researchers for tasks including model training, benchmarking, and perform model tuning.