allenye66's Stars
simonmeusel/MuteSpotifyAds
A efficent MacOS application automatically silencing ads of the spotify desktop app
allenye66/Computer-Vision-Lip-Reading-2.0
A speech recognition system using 3D CNNs. The final model achieves 97.4% training accuracy and a 99.2% testing accuracy and the system can accurately recognize spoken words from a set of pre-defined words in real-time.
allenye66/Deep-Learning-Autonomous-Drones
Here is a conglomeration of file depcting the code we wrote to create an autonomous drone using a CNN-LSTM model to aid in food and package delivery during the 2020 quarantine. Steering Angle Dataset Exploration: Here is where we explored methods in making our CNN-LSTM predictor, as well as coded the final version. We also have graphs for the results of our code. We also define the Gaussian and Edge detection preprocessing functions over here. Yolov3 Bounding Boxes: Here is where we created a transfer learning model from the Yolov3 architecture to find bounding boxes of cars, people, and trees in our images. These bounding boxes were used by our probability model to calculate the probability of collision. Weight determination functions: Here is where we defined the functions user to calculate the probability of colliding into any given object. The final probability determination function can be found in the Yolov3 script, as well as the UserModelLibrary scripts. Data Exploration: Here is where we explored the data intially given to us, and found that the data was abnormally distributed. This helped us deermine the wraparound problem, as well as why our models prediction were near 0 in the early stages of the process. Trial.py: This is the script to fly the actualy drone. UserModelLibraries: This is the final conglomeration of all of our code - the probability functions, models, and pre/post processing function used to run our algorithms. All pictures and graphs are also included in the pictures and graphs photo.
allenye66/HandwritingToLatex
LaTeX is a great tool for authors of research papers to format and typeset math formulas, we wanted to try working with LaTeX and converting handwriting to LaTeX code. How this project works is by taking a JavaScript canvas image and sending it to a Flask server. This server then runs the image through a convolutional neural network, trying its best to predict the numbers and characters and finally converts the expression into LaTeX form.
allenye66/Computer-Vision-Sudoku-Solver
How our program works: After the user submit an image, we send the image to our flask server. From there, we preprocess the image using Gaussian Blurring and Thresholding. Afterwards, we locate the sudoku grid in the image using Hough Line Transform algorithm. After having an image with just the grid, we splie the grid into 81 seperate images and use a CNN with approximately 99 percent accuracy to predict the numbers in each grid. These numbers are added to an array and we use our sudoku solving algorithm to solve the puzzle.
allenye66/allen-ye.com
allen-ye.com
allenye66/Connect4-With-Minimax
Connect-4 game with AI of different difficulties to play against. Also supports single player.
allenye66/Integraph
Easy integration
WebifyBayArea/Webify
A website development business to help local businesses and restuarants during COVID-19