Object Detection in High Resolution Image

Using high-resolution satellite images as training data, developed a KNN algorithm to detect red cars in the picture by using RGB pixels. KNN was implemented from scratch.

Close Window

Supervised Learning

Developed KNN, probabilistic generative function to classify three different datasets with 2, 7, and more than seven dimensions. Both algorithms have been implemented from scratch. For hyperspectral data, the diagonal probabilistic generative method has been implemented to avoid the problem of a singularity of the covariance matrix.

Close Window

English Handwritten Character Recognition

After collecting the data, including English handwritten characters, multiple steps have been applied to extract features from data. In the next step, Knn method has been applied to classify the characters. Multiple experiments have been conducted to investigate the impact of hyperparameters such as k and train/test split size and distance metric on the classification accuracy of the underlying method.

Close Window

US Visa Application

Developed clustering methods such as divisive, hierarchical, and partitioning on ”US Permanent Visa Applications” data from Kaggle. We investigated which majors would get more approval for worker visa and visa approval rate based on country. Developed classification methods such as naive bays, decision tree, random forest, svm on ”US Permanent Visa Applications” data from Kaggle. We predicted the visa status by considering other features.

Close Window

Arrhythmia Detection

• Developed Linear Discriminative Anaylysis and Quadratic Discriminative Analysis on ” Arrhythmia Data Set” from UCI ML repository.

Close Window

Classification on laser induced Plasma spectrum

Using laser spectrum as our input data, we are working on deep neural network problem to classify different materials exist in the spectrum.

Close Window