Inspiration
In our modern world, Big Data and Machine Learning have been buzz words that spark interest in innovation. Initially, we were inspired to try to understand machine learning more, and use its applications to determine if a car is stolen or not, but after experimentation, we were able to see that there is quite literally too much data. However, this did not stop us, we only slightly changed our viewpoints. We instead focused our attention on creating a way to pre-process a license-plate and splice it, that way we can feed it into a machine learning algorithm that suits our needs.
What it does
Our License Plate Pre-Processing is a program that uses our ideas of how to get a license plate pre-processed into something that a Cnn could use. By applying grayscale, and binarization, an image of a license plate becomes only slightly closer to having it capable of being sent as an input. Our program does the entire process of extracting an image of a license plate from a folder, apply the grayscale and binarization, and finally using a process of splicing each character of an image out.
How we built it
We built the License Plate Pre-Processing with the idea that the majority of the time will be spent researching how to fix something or looking up functions. Our program was built in Python and used Keras, numpy, cv2, etc. The entire process was like a jigsaw puzzle where we would look up how to do something, and glue pieces together until we could see the full picture.
Challenges we ran into
The biggest challenge we faced was not knowing anything about what we threw ourselves into, however in its way it made it somewhat more enjoyable. It is because of these setbacks that it felt so good to create a few lines of code that would solve an error that we were clueless of its origins.
Accomplishments that we're proud of
We are proud of our curiosity to try new things. Also, we are proud of our thought process, if we could not figure out how to separate characters of license plates into separate images, we make it our project to make a path. Finally, we are proud of our commitment to attempt something harder.
What we learned
Just about everything that we tried this weekend was brand new to us. For this reason, it was a learning experience to see how important modules are and the different ways you can download them, (pip, conda, etc.) Besides, It broadened my point of view about machine learning and its related fields, as well as opening my eyes to different libraries that assist in creating a Neural network.
What's next for License Plate Pre-Processing
Next up on the List for License Plate Pre-Processing is using the program to splice license plates and use it to create a CNN. Furthermore, we would like to complete our idea of making the CNN return a string of a car's license plate where we can determine if a vehicle is stolen or not.
Log in or sign up for Devpost to join the conversation.