Fruit Image Classifier (TensorFlow & Keras)

This project is a Convolutional Neural Network (CNN) built with Python and TensorFlow/Keras to classify images of different fruits. It features a split architecture: one script for training and saving the model, and another script with a simple GUI to select an image and make predictions instantly.

Features

• Custom CNN Architecture: Uses multiple Convolutional and MaxPooling layers to extract features from images.
• Separated Workflows: Train the model once and use the saved model for endless predictions without retraining.
• Simple GUI: Uses tkinter to open a native file dialog box, making it easy to select test images.
• Performance Graphs: Automatically plots training vs. validation accuracy using matplotlib.
• GPU Support: Configured to leverage NVIDIA GPUs for faster training (if available).

📁 Project Structure

📦 Your-Repository-Name
 ┣ 📂 trainning/           # Your training images (organized by fruit subfolders)
 ┣ 📂 test/                # Your validation/test images (organized by fruit subfolders)
 ┣ 📜 projectfruit.py      # The script to train the model and save it
 ┣ 📜 result.py            # The script to load the model and predict a selected image
 ┣ 📜 fruit_model.keras    # The saved model (generated after running projectfruit.py)
 ┗ 📜 class_names.json     # The saved class labels (generated after running projectfruit.py)

🛠️ Prerequisites Make sure you have Python installed, along with the following libraries:

pip install tensorflow matplotlib numpy

How to Use

1. Prepare Your Data Ensure you have your dataset organized into trainning and test folders inside the project directory. Inside these folders, create a subfolder for each fruit (e.g., trainning/Apple, trainning/Banana, etc.).

2. Train the Model Run the training script. This will process your images, train the CNN, display an accuracy graph, and save both the model (fruit_model.keras) and the class names (class_names.json). python projectfruit.py

Model Details

The neural network is built using the Sequential API in Keras and includes:

4 Conv2D layers with ReLU activation for feature extraction.

4 MaxPooling2D layers to reduce spatial dimensions.

A Flatten layer to convert the 2D matrices into a 1D vector.

A Dense hidden layer with 512 neurons.

A Dropout layer (0.5) to prevent overfitting.

A final Dense output layer using the Softmax activation function to determine the most likely fruit class.