USC CSCI544 - Applied Natural Language Processing - Fall 2023 - Prof Mohammad Rostami

『ゼロから作る Deep Learning ❸』(O'Reilly Japan, 2020)

This Python Project code implements a simple neural network built from scratch using the NumPy library to classify whether a given number is odd or even.

📚 A simple PyTorch nueral network API capable of generating new text based on starting text given trained using Discord Messages

Q&A Recognition AI Template

Coloring images with a Convolutional Autoencoder

Stock Market Direction Predictor

TMNN: A Python library for building and experimenting with custom neural networks from scratch

New Article Analysis using KNN, Logistic Regression, Neural Networks and Multinomial Naive Bayes in python

Google images scrapper in full resolution with multi threading speed

Waste Management based project, uses a fast real-time object detector to classify and localize objects present in a live feed through a webcam, which will help in segregation. It can also be used in autonomous surveillance during ban.

The code here can be used to train a Transformer Neural Network to perform symbol recovery at the receiver end.

MITx micromaster class, Machine Learning w/ python, from linear models to deep learning

Sign language gesture recognition is done in two ways. Alphabets are detected from sign language and words are formed using this. The words thus formed is then converted to speech. Another method includes recognizing gestures which include words.

Implement intelligent agent for minigames using neural network and evolution algorithm

simple chat bot which advices users specific cure based on their symptoms

The project is a neural network-based fashion recommendation system built using Python. The model used for this system is Resnet50, which is a deep learning model used for image recognition. The data used for training the model is scraped from Flipkart, with a total of 65,000 images.

Music Generation using Deep Learning

Visual Assistant is an interface which will give the scene understanding to its users by using machine learning and image processing algorithms on live video.

4th place submission out of 20 teams (undergrad + masters + phd competitors)