Hareesh Bahuleyan

Image Super Resolution [Code]

A CNN architechture (with residual connections) is trained to transform low resolution images to high resolution. The input to the model is a 72x72 image which is fed through a series of convolutional and residual blocks (to learn necessary features), and then upsampled (either by UpSampling2D or Deconv2D from Keras) to obtain a 288x288 image.

Neural Style Transfer [Code]

Given two images A (content image) and B (style image), the task is to transfer the style of image A into that of image B, while being able to preserve the content of image A. The idea is to start with a random noise image (input image), and to update its pixel values so that we minimize both the style and content loss.

Music Genre Classification [Code]

Automatic recognition of music genre using the ensemble of two approaches: (1) Spectrogram based end-to-end image classification using a CNN (VGG-16), (2) Feature Engineering Approach using Logistic Regression, SVMs, Random Forest and eXtreme Gradient Boosting.

Music Language Modelling [Code]

Generate music using language modelling approach with LSTM neural networks. MIDI instructions are converted into a sequence of 'words' and the task is to predict the next word in the sequence, given the previous _n_ words.

Kaggle Seizure Prediction Challenge [Code]

EEG signals record electrical activity of the brain and can be used to predict the onset of a seizure. A competition conducted by the University of Melbourne on Kaggle required participants to classify 10-minute EEG clips into clips into either interictal or pre-ictal. I appoached the problem using an XGBoost classifier trained on (1) Time domain features, (2) Frequency domain features, and (3) EEG Specific Features.

Text Classification using LSTMs [Code]

Carry out a sentiment analysis task on the Rotten Tomatoes movie review dataset using recurrent neural networks. Specifically, an LSTM was trained with sentences (sequence of words) to predict the sentiment label.

Airbnb Rental Price Prediction [Code]

Python Selenium is used to scrape listings data from airbnb.ca for the city of Toronto. An extensive exploratory data analysis is carried out, following which, a price prediction model is built using information such as in-house amenitites and neighbourhood facilites as features.