Teaching Assistant, Computer Science resume example

Publications:

Research Projects

American Sign Language Classification:

The project aimed to use SVM, CNN among other machine learning models to create an image classification system for ASL. The ASL dataset was first gathered, and the photos were then preprocessed using scaling, normalization, and data sampling. Models were then trained and tested using the preprocessed images as input, mainly PCA as a preprocessing technique. For model performance, we looked at the accuracy of each model against the execution time taken by each of them. According to the findings, the CNN model with two layers had an accuracy of 73.76%, while the SVM model with PCA had a maximum accuracy of 86.42%. In comparison to the other models, the SVM model with PCA also had the quickest execution time. The computational complexity of training and testing deep learning models on huge datasets was one of the main difficulties encountered, and hence we performed sampling.

Chest cancer classification

The project aimed at classifying chest X-ray images to detect the type of chest cancer. The images were preprocessed through augmentation and regularization. Models implied were CNN and ResNet50V2. Transfer learning was performed to increase the accuracy of model predictions.

Food Image Classification

Food recognition has become an important task in computer
vision due to the increasing interest in developing intelligent
systems for meal planning, calorie tracking, and dietary analysis
ysis. Accurately identifying food items in images is crucial.
for these applications. To address this, we proposed a ma-
chine learning model that can precisely classify food items
from the images. The dataset contains static images of food
items with 11 classes, and each class has around 100 images.
for the same food item. The models we aimed at was the clas-
sic CNN model for image processing and also transfer learn-
ing by the EfficientNet model. To improve the model’s per-
formance, we applied preprocessing techniques such as sam-
pling, resizing, and Principal Component Analysis (PCA) for
feature extraction. The EfficientNet model with preprocessed
data performed best in terms of accuracy and computation
cost. This efficient, model has the potential to aid proper nu-
trition diet to users.