Harshitadd's Projects
Runner's Up submission for the Singapore India Hackathon - Developed a panoramic object detection system - that avoids the detection of the same object ( which is common in the n pictures of a panoramic view ) using image stitching. This was used with a static recommender to detect the possible actions that can be taken by a user to mitigate the effects of high Volatile Organic compounds in any object. This system when used in conjunction with a handheld VOC detection device and a semi-autonomous rover could be used as an effective solution to tackle high VOC concentration identification and reduction even in unattended environments.
Finalist ACM SRC 2020 @ASE: This repository contains our work submitted under the ACM Student Research Competition of The 35th IEEE/ACM International Conference on Automated Software Engineering. The published short paper will be linked shortly.
TFDS is a collection of datasets ready to use with TensorFlow
Analysing the carbon footprint of Differential privacy guarantees on Vanilla BERT ( modelled for news classification ), MNIST and the Cartpole Problem using Opacus. [ Accepted to appear at ICML 2021 - Workshop on Socially Responsible Machine Learning ]
Low-Resource OCR
Scripts for indicTranslate - Machine Translation for Indian languages
Contributions to AmericasNLP Shared Task for Low-Resource ASR
Just a plain, simple and elegant one-page theme for research/academia.
Official Code for Submission to SemEval's Shared Task 11 for Propaganda Detection
Repository for SemEval 2020 Task 11 related topics
🎓 Easily create a beautiful academic résumé or educational website using Hugo, GitHub, and Netlify
Winning submission in eYantra 2019 : A LFR which engages a custom built electromagnetic arm mechanism upon the identification of an appropriate aruco marker. The project uses path planning with python and OpenGL,Blender animative elements as well.
Runner's Up submission under The Marconi Society's Celestini Prize India. This repository entails my contributions to our project VisionAir : A privacy-preserving smartphone application that can be used to estimate the Air Quality Index from an image that the user takes. The application uses On-Device training to ensure complete isolation of the user's private data and also uses federated learning to diversify the response of the global deep learning model iteratively