Camera Classifier with live video feed,
custom Tkinter GUI, and CNN model trained on user images.
Camera Classifier is a Python application that allows users to capture images from a live camera, label them, train a Convolutional Neural Network (CNN) model, and predict classes in real-time.
The project combines a CustomTkinter GUI, OpenCV camera handling, and a CNN built with TensorFlow/Keras, providing a complete interactive experience for image classification.
Key features:
- Custom class setup: The user chooses the number of classes (2–10) and assigns their names.
- Image collection: Captured images are automatically stored in numbered folders (e.g.,
1,2,3, …) corresponding to the classes. - CNN training: Images are preprocessed and augmented. Since this is a CNN, training may take some time if many images are collected. During training, the video feed will freeze: this is normal and helps optimize performance.
- Real-time prediction: The model can predict classes from the live camera feed using either the Prediction or Auto Prediction modes. These functions are quite sensitive, so accurate results require a large and balanced image set. For best performance, try to maintain consistent lighting, environment, and camera distance.
- Reset functionality: Clears the model and deletes all saved images, but keeps the empty class folders. These can be removed manually if not needed.
This tool is ideal for small-scale image classification experiments, live demonstrations, and learning CNN fundamentals.
- Download the latest CameraClassifier folder (includes all files and the batch file).
- Make sure Python 3.x is installed and added to your system PATH.
- Open the folder and double-click
run_camera_classifier.batto start the application. On first run, the batch file will automatically install all required Python packages (fromrequirements.txt).
- The program prompts for class number and names.
- Captured images are stored in numbered folders (e.g.,
1,2, …) that match the chosen classes. - Train the CNN model with the collected images. Note: while training, the application will temporarily freeze and the live image will stop updating until training is completed.
- Use the "Predict" or "Auto Prediction" buttons to classify the live camera feed.
- The "Reset" button deletes all collected images but leaves the folders intact (they can be removed manually if desired).
- The status bar and labels display prediction results in real-time.
Released under the MIT License. Feel free to use, modify, and share 🚀