Skip to content

Latest commit

 

History

History
89 lines (71 loc) · 3.69 KB

File metadata and controls

89 lines (71 loc) · 3.69 KB

Grapevines Species Classification Project

A. Project Overview

  • Grapevines are significant agricultural plants, primarily cultivated for their fruits, which are consumed fresh or processed into various products such as wine, juice, and raisins. Additionally, grapevine leaves hold importance as they are harvested annually and used for culinary purposes, particularly in Mediterranean cuisine. The species of grapevine leaves play a crucial role in determining the quality, taste, and market value of the final products.

B. Objective

  • The objective of this project is hosted as a Kaggle competition, to develop a deep learning-based classification system capable of accurately identifying grapevine leaf species from image data. By leveraging the power of deep learning algorithms, we aim to overcome the limitations of manual classification and provide a reliable and efficient solution for grapevine leaf species classification. This project follows a systematic approach, encompassing the following steps:

    1. Data Collection: Obtain a dataset consisting of images of grapevine leaves belonging to different species.

    2. Preprocessing: Apply preprocessing techniques such as resizing, normalization, and augmentation to prepare the data for training.

    3. Model Selection: Choose appropriate deep learning CNN architectures for the classification task, considering factors such as model complexity and computational resources.

    4. Model Training: Train the selected models using the preprocessed dataset, optimizing model parameters to minimize the classification error.

    5. Model Evaluation: Evaluate the trained models on a separate test dataset to assess their performance in accurately classifying grapevine leaf species.

    6. Performance Analysis: Analyze the performance of the trained models using metrics such as accuracy, precision, recall, and F1-score.

    7. Fine-tuning and Optimization: Fine-tune the models and explore optimization techniques to further improve classification performance.

    8. Deployment: Deploy the trained classification system for practical use, potentially integrating it into agricultural processes for automated species identification.

Methodology

  • In our exploration of deep learning architectures for grapevine leaf species classification, we initially experimented with building custom Convolutional Neural Network (CNN) architectures from scratch. However, these attempts yielded unsatisfactory results, with the models struggling to predict and classify the images correctly, resulting in low accuracy. Recognizing the limitations of our custom architectures, we transitioned to leveraging pretrained CNN architectures, which have been trained on large-scale image datasets and have demonstrated strong performance in various image classification tasks. We explored several state-of-the-art pretrained architectures, including Inception, EfficientNet, MobileNet, Xception, Inception-ResNet and ViT (Vision Transformer), adapting them to suit the requirements of our grapevine leaf species classification task.

Model Deployment

Model.Deployment.mp4

Credits: