FruitSeg30_Segmentation dataset & mask annotations: A novel dataset for diverse fruit segmentation and classification

Area of Science:

• Computer Vision
• Agricultural Technology
• Deep Learning

Background:

• Fruits are vital for human nutrition, necessitating efficient agricultural processes.
• Accurate fruit classification and segmentation are critical for automated sorting and quality control, reducing costs and enhancing consistency.
• Existing datasets may lack the diversity and quality required for robust deep learning models in fruit analysis.

Purpose of the Study:

• To introduce the FruitSeg30_Segmentation Dataset & Mask Annotations, a novel dataset for fruit segmentation and classification.
• To provide a diverse and high-quality image collection for training advanced deep learning models.
• To establish new benchmarks in dataset quality and diversity for agricultural applications.

Main Methods:

• Development of the FruitSeg30 dataset with 1969 images across 30 fruit classes.
• Utilizing a U-Net architecture for deep learning model training.
• Performance evaluation using metrics such as accuracy, precision, recall, F1-score, IoU, and Dice score.

Main Results:

• The U-Net model achieved high performance metrics: 94.72% training accuracy, 92.57% validation accuracy, 94% precision, 91% recall, 92.5% F1-score, 86% IoU, and a 0.9472 Dice score.
• The dataset demonstrated superior performance in segmentation tasks.
• The results indicate the effectiveness of the FruitSeg30 dataset in advancing fruit segmentation capabilities.

Conclusions:

• The FruitSeg30 dataset addresses a critical gap in resources for fruit image analysis.
• The dataset enhances the potential of deep learning models in agricultural technology and the food industry.
• This work sets new standards for dataset quality and diversity in fruit-related AI applications.