Bark Texture Recognition of Indian Trees: A Bangalore-Centric Dataset and Analysis
Description
This dataset focuses on the bark textures of 22 tree species native to the Karnataka region, specifically Bangalore, India. The images were captured from various angles, including front-facing and distant views, ensuring a diverse representation of bark textures across different tree maturity stages, from young saplings to fully mature trees. This dataset is particularly useful for a wide range of applications, including timber industry analysis, ecological studies, and advancing machine learning models for tree species classification. The variety in angles, lighting conditions, and growth stages makes the dataset versatile for training robust bark texture recognition systems.
Files
Steps to reproduce
To gather the data for this tree bark texture classification project, we followed a systematic approach to ensure diversity, consistency, and quality, aiming to create a robust dataset for model training and evaluation. 1. Data Collection: The data was gathered from 25–30-year-old trees in Karnataka, India, spanning various geographical regions, including Bangalore. The trees were chosen to ensure diversity in species, growth stages, and environmental conditions. We aimed to capture images from different tree species to fill the gap in existing datasets, which lacked Indian-based tree species for bark texture classification. 2. Image Capture Protocol: Images were captured using a mobile phone camera, chosen for its portability and ability to capture real-time data in diverse natural settings. The images were taken in varying environmental conditions, ensuring that the dataset reflected different lighting, background, and tree maturity levels. We aimed to capture a variety of tree bark textures, including smooth, rough, and fissured patterns, from both front-facing and side angles, along with some distant views to provide a complete representation of bark texture across species. We considered lighting conditions (e.g., bright sunlight, overcast) and background diversity to ensure the images represented real-world scenarios. 3. Data Preprocessing: The raw images were converted to JPEG format for consistency and compatibility with standard image processing tools. The images were resized to a consistent resolution to ensure uniformity, and augmentation techniques, including rotation, flipping, and scaling, were applied to increase dataset variability and prevent overfitting. Image annotations were manually done for each tree species, ensuring that the dataset could be used for supervised learning tasks. 4. Software and Tools: Various software tools were used throughout the data collection, preprocessing, and model development stages. Image editing and augmentation were done using Python libraries like OpenCV and PIL. The dataset was organized and labeled using a custom Python script, ensuring that each image was correctly associated with its respective tree species and environmental conditions. Additionally, tools like Jupyter Notebooks and TensorFlow/Keras were used to design and evaluate machine learning models. 5. Reproducibility: To ensure the reproducibility of our work, detailed protocols were followed in data collection, preprocessing, and model evaluation. Scripts for image preprocessing and augmentation are made publicly available, along with the dataset and model training configurations. By sharing this information, we aim to help others replicate the data collection process and experiment with similar tree bark texture classification challenges. This process created a robust dataset for advancing tree bark texture classification research.