Milestone Report 5: Traffic Sign Recognition

GROUP: G
TEAM MEMBERS:
Jasleen Kaur (C0886477)
Komal Astawala (C0883392)
Mayankkumar Patel (C0883127)
Shree Prada Devaraju (C0887257)
Simran (C0896133)

 Introduction
This report presents the advancements achieved in the project focusing on 'Traffic Sign
Recognition utilizing Convolutional Neural Networks (CNNs).' It primarily emphasizes the
exploration of diverse CNN architectures tailored specifically for this task, with meticulous analysis
of their fundamental characteristics. Additionally, notable updates include the integration of
saving the trained model as an .h5 file format, along with the incorporation of Tkinter for the
development of a user-friendly interface, enhancing accessibility and usability of the model.

 Custom TSRModel
Our decision to use the custom CNN model for Traffic Sign Recognition (TSR) is driven by its
superior performance on both training and test datasets, surpassing pre-trained models available
in the market. Our model, initialized as `TSRModel`, is meticulously designed to capture intricate
features present in traffic sign images. Incorporating convolutional layers for feature detection,
ReLU activation functions for non-linearity, pooling layers for downsampling, and Dropout layers
for mitigating overfitting, our model demonstrates robustness and generalization capabilities. The
sequential organization of layers simplifies model construction and modification, facilitating
experimentation with different architectures and hyperparameters. Additionally, the model is
compiled with appropriate settings, enhancing its effectiveness for multi-class classification in the
TSR domain.

 Utilizing .h5 Format for Storing Trained Models: Enhancing Model Persistence and
Portability
Using .h5 format for storing trained models offers benefits in versatility, efficiency, and portability.
It supports various data types, provides efficient storage and compression, and ensures
compatibility across different platforms and frameworks. This makes it a convenient choice for
managing large models and facilitating easy sharing and deployment.

 Enhancing User Interaction: Implementing a Graphical User Interface (GUI) for Traffic
Sign Recognition
 Tkinter is a standard Python library for creating graphical user interfaces (GUIs). It provides a
simple and easy-to-use framework for developing desktop applications with interactive
elements such as buttons, labels, and entry fields.
 Benefits of Using Tkinter:
1. Ease of Use: Tkinter is beginner-friendly and easy to learn, making it accessible to
developers of all skill levels.
2. Cross-Platform Compatibility: Tkinter applications can run on various operating systems
without modification, ensuring broad compatibility.
 3. Integration with Python: Tkinter seamlessly integrates with Python, allowing developers
to leverage Python's rich ecosystem of libraries and tools.
4. Built-in Widgets: Tkinter provides a wide range of built-in widgets for creating interactive
GUI elements such as buttons, labels, and entry fields.
5. Customization: Tkinter allows for extensive customization of GUI elements, including
colors, fonts, and styles, enabling developers to create visually appealing interfaces.
 Features of the GUI Application:
1. Window Configuration: The application window (top) is configured with a specific size,
title, and background color.
2. Image Upload: Users can upload an image using the "Upload an image" button, which
opens a file dialog for selecting an image file.
3. Image Display: The uploaded image is displayed in the GUI using a Label widget
(sign_image).
4. Classification: Users can classify the uploaded image by clicking the "Classify Image"
button, which triggers a classification process based on a pre-trained model.
5. Result Display: The predicted class label for the uploaded image is displayed below the
image using another Label widget (label).
6. GUI Layout: Widgets are arranged using various Tkinter layout managers (pack, place) to
achieve the desired layout and positioning.
 Potential Improvements:
1. Enhanced User Feedback: Provide additional visual feedback to the user during the
classification process, such as progress indicators or status messages.
2. Error Handling: Implement robust error handling to gracefully handle exceptions or
unexpected user inputs.
3. Improved Layout and Design: Enhance the visual appearance of the GUI by refining the
layout, adding spacing, and applying consistent styling.
4. Interactive Elements: Incorporate interactive elements such as sliders, checkboxes, or
dropdown menus to enhance user interaction and functionality.
5. Performance Optimization: Optimize the classification process for speed and efficiency,
especially for handling larger images or complex models.
6. Localization and Internationalization: Support multiple languages and locales to cater to
a global audience.
7. By incorporating these improvements, the GUI application can become more user-
friendly, visually appealing, and functional, providing a better overall user experience.
Additionally, leveraging Tkinter's flexibility and ease of use allows for quick iteration and
experimentation to refine the GUI further.
 User Interface