banana_ripening_dataset_day0_to_day7

Description

This dataset presents a structured time-series collection of banana images captured over a period of eight consecutive days (Day 0 to Day 7) to analyze the visual progression of fruit ripening. The dataset was developed as part of a research internship project focused on applying computer vision and machine learning techniques to agricultural and food quality assessment. The dataset consists of approximately 30 individual banana samples, where each sample is uniquely identified and stored in separate folders (Banana_ID_xxx). Each folder contains a sequence of images (day0 to day7) representing the same banana observed daily under controlled and consistent environmental conditions. This organization enables accurate tracking of physical and color changes associated with the ripening process over time. The dataset is suitable for applications such as image classification, ripeness stage detection, and time-series analysis. All images are captured with consistent background settings to minimize noise and ensure reliable feature extraction. The dataset structure supports both per-sample temporal analysis and cross-sample comparison. This dataset can serve as a benchmark resource for researchers and practitioners working in computer vision, agriculture, and food quality assessment domains.

Steps to reproduce

## Steps to Reproduce 1. Download the dataset ZIP file and extract it to a local directory. 2. Ensure the dataset is organized as follows: * Each folder (Banana_ID_xxx) represents a single banana sample. * Each folder contains sequential images from day0.jpg to day7.jpg. 3. (Optional) Create or use a metadata file mapping image paths to banana ID, day index, and ripeness label. 4. Preprocess the images: * Resize images to a fixed resolution (e.g., 128x128 or 224x224 pixels). * Normalize pixel values (e.g., scale to range [0,1]). * Apply data augmentation if required (rotation, flipping, brightness adjustment). 5. For classification tasks: * Assign labels based on ripeness stage: * Day 0–2: Unripe * Day 3–5: Semi-ripe * Day 6–7: Ripe 6. Split the dataset into training, validation, and testing sets (e.g., 70:15:15 ratio), ensuring that images from the same Banana_ID are not split across different sets for time-series consistency. 7. Train a machine learning or deep learning model: * Example models: CNN, ResNet, or time-series models. * Frameworks: TensorFlow, Keras, or PyTorch. 8. Evaluate model performance using appropriate metrics such as accuracy, precision, recall, and F1-score. 9. For time-series analysis: * Use sequential models (e.g., LSTM or CNN-LSTM) to capture temporal dependencies across days. 10. Visualize results: * Plot accuracy/loss curves. * Analyze prediction results across different ripening stages. These steps provide a general guideline for reproducing experiments and building machine learning models using the dataset.

Institutions

• Savitribai Phule Pune University

  Maharashtra, Pune

Categories

Licence