Inspiration

Sign Mapper is a tool developed to aid drivers and autonomous vehicles in navigating roads safely and efficiently. The inspiration behind the system was the need for accurate and reliable systems for detecting and interpreting road signs. With the increasing prevalence of self-driving cars, it has become crucial to have such systems in place. Sign Mapper uses deep learning techniques, specifically a convolutional neural network, to classify and locate different types of signs, including stop signs, yield signs, and speed limit signs, among others. This deep learning model allows for a high degree of accuracy in recognizing and interpreting signs, making it a valuable tool for ensuring road safety.

What it does

Sign Mapper is a novel web-based application that facilitates the automatic classification of road signs using a convolutional neural network (CNN). This CNN model has been trained on a vast collection of road sign images, which enables it to recognize and classify diverse types of signs, including, but not limited to, stop signs, yield signs, and speed limit signs. The application employs a user-friendly interface where users can upload a road sign image, and the system outputs a visual prediction of the road sign type with an accompanying name and identification number. This advanced feature ensures precise and accurate identification of road signs, promoting road safety and navigation efficiency.

How we built it

Sign Mapper was constructed using the Streamlit framework in conjunction with Python. The Keras deep learning library was utilized to develop the model, which was subsequently trained on an extensive collection of over road sign images. The development process involved the creation of three models, including Convolutional Neural Networks (CNN), Support Vector Machines (SVM), and decision trees. Notably, the CNN model yielded the highest accuracy of 97.4%; hence, it was employed in the construction of the Streamlit application. The application was developed using interactive widgets and visualizations provided by the Streamlit framework and deployed on a cloud platform to enhance accessibility and convenience.

Challenges we ran into

Acquiring and cleaning the road sign image dataset presented one of the most significant obstacles during the development of Sign Mapper. Additionally, issues related to class imbalance and overfitting emerged during the model training process. These challenges necessitated meticulous tuning of the model architecture and hyperparameters to ensure optimal performance.

Accomplishments that we're proud of

Our team takes pride in the impressive accuracy and resilience of our model. It effectively identifies and classifies an extensive array of road signs with remarkable precision and recall. Furthermore, we are delighted with the user-friendly interface that we designed using Streamlit. The intuitive interface allows users to effortlessly add image links and receive real-time predictions, ultimately facilitating efficient and straightforward road navigation.

What we learned

The development of Sign Mapper provided valuable insights into the complexities and opportunities associated with building machine learning applications for real-world challenges. Our team gained extensive experience in various tools and techniques, including transfer learning, data preprocessing, and Streamlit, among others. This practical experience enabled us to hone our skills and enhance our understanding of best practices in developing effective and efficient machine learning applications.

What's next for Sign Mapper

Our team recognizes the significance of our work on Sign Mapper and the implications it has for real-world challenges. In the future, we aim to advance this technology and improve the accuracy and functionality of the model. By expanding the range of supported road signs and integrating with other navigation and mapping tools, we believe Sign Mapper can have a significant impact on improving road safety and reducing accidents. Furthermore, our efforts to explore novel techniques for image segmentation and localization have the potential to enhance the accuracy and efficiency of Sign Mapper, providing a more reliable and comprehensive solution for road sign detection and classification. Overall, Sign Mapper is a promising tool with substantial implications for the real world, and we are excited to continue working on its development.

Challenges of the current model

The current version of Sign Mapper has a limited number of supported road signs and a dataset of 42 images. As a result, there may be instances where the model fails to accurately classify an image, resulting in an invalid output. Our team is aware of this limitation and is continuously working to expand the dataset and incorporate support for additional types of road signs. This ongoing effort will improve the accuracy and robustness of the model, ultimately increasing its effectiveness as a tool for safe and efficient road navigation. We appreciate the feedback and will continue to work towards improving the performance of Sign Mapper.

Usage

Sign Mapper has the potential to benefit both autonomous self-driving cars and drivers with vision problems. For self-driving cars, Sign Mapper can serve as an essential tool for accurately detecting and interpreting road signs, which is crucial for safe and efficient navigation. With the increasing prevalence of autonomous vehicles on the road, it is important to have accurate and reliable systems in place for detecting and interpreting road signs. Sign Mapper's deep learning-based classification and localization techniques can aid in the development of such systems, providing a more efficient and reliable way for autonomous vehicles to navigate the roads.

For drivers with vision problems, Sign Mapper can serve as a useful aid in identifying road signs that they may have difficulty seeing. By using Sign Mapper to take a picture of a road sign, a driver can receive an accurate classification of the sign's content, enabling them to better navigate the road. This can be especially helpful for those with visual impairments or conditions such as color blindness, where traditional road sign recognition may be challenging. Overall, Sign Mapper has the potential to improve road safety and efficiency for a wide range of drivers, including those with visual impairments and autonomous self-driving cars.

Built With

• cnn
• decision-tree
• python
• streamlit
• svm