This was my entry for Grab's AI for S.E.A. challenge. This repository includes the notebooks and notes on the challenge approach. Selected as one of the Top 10 entries for the competition. Used and ensemble of SARIMA and LSTM, though a Random Forest Regressor, for the forecasting model . Pandas and Matplotlib were used for the data exploration.

Trained a text classifier to predict the sentiment based on the input text. Using spaCy, I have created a lightweight model that can classify a text as either positive or negative. This was trained using the IMDB review dataset. Want to check a simple demo? Click on the icon above.

What I have done was to implement a Deep Deterministic Policy Gradient (DDPG) model as a controller for a quadcopter. The objective I chose was for the agent to move and hover to a position defined. I have created a reward function that accounts for the velocity and relative distance between the agent and the objective.

A fork on the implementation of the paper Super SloMo: High-Quality Estimation of Multiple Intermediate Frames for Video Interpolation. The model in this script can convert video inputs into multiple interpolated frame outputs which can then be compressed into framerates to make the output appear slow-moving or increase the output framerate from the original.

The goal of this project is to showcase what we have learned in the first phase of the challenge which covered an introduction to PyTorch, PyTorch basics and the different deep learning models with coding in PyTorch. For this project, we are asked to create a classifier model that would be able to output the predicted category of a flower among 102 categories.

This project is an implementation of an RNN that can continue creating a script based on the original input text. The sample script for this project is on a scene from The Simpsons series. The generated text showed some unusual results and made a scene that appeared to have continuity from the original script.