Vegetable Classification using Transfer Learning with InceptionV3
-
Overview: A comprehensive exploration of vegetable classification using transfer learning with the InceptionV3 architecture. Developed in a Colab environment for ease of accessibility and collaborative use.
-
Training Notebook: Guides users through the complete pipeline: dataset loading, preprocessing, EDA, model construction, and evaluation. Incorporates image enhancement techniques, data visualization, and fine-tuning of the InceptionV3 model. Utilizes TensorFlow and Keras for efficient deep learning implementation.
-
Inference Notebook: Demonstrates the model's inference capabilities on new vegetable images. Loads a pre-trained InceptionV3 model and applies image preprocessing, including color and contrast enhancement. Showcases the model's adaptability to different input conditions.
-
Dataset and Model: The training dataset comprises 15 distinct vegetable classes, ensuring diversity and representation. Model files and visualizations, including accuracy and loss curves, are included in the repository. DataSet link: https://www.kaggle.com/datasets/misrakahmed/vegetable-image-dataset
-
Educational Resource: Serves as an educational tool for those interested in deep learning, transfer learning, and image classification tasks. Colab notebooks facilitate replication of the training process and exploration of model performance.
-
GitHub Repository: Detailed documentation for easy replication and understanding of the project. Tools for model evaluation, including a confusion matrix and visualizations of misclassified images.
-
Practical Insights: Offers practical insights into the entire lifecycle of a deep learning project – from data preparation to model deployment. Valuable for individuals seeking hands-on experience in transfer learning for image classification tasks. This project combines educational and practical elements, making it an insightful resource for both learners and practitioners in the field of deep learning and computer vision.