Projects
Aesthetic-Face-De-Identification
A scalable approach protect the privacy of people in the background of the frame via a face de-identification filter
Tensorflow OpenCV Android Image Filter
A scalable approach to preserving the privacy of bystanders in social media posts :
- Entirely ‘edge’ system tested on low end android phones-
- Client side nature makes it scalable as no server compute required for new nodes,
- Edge compute improves privacy as unfiltered data never leaves the device
- Aesthetics are preserved via custom filter implementation based on human tendecies
ARCNN-keras
A tf-keras implementation of ARCNN
Tensorflow Keras Docker tf.data PIL
A tf-keras implementation of ARCNN mentioned in :
- Implemented ARCNN in tf.keras for easy of use
- Created a novel “lite” version using dilated convolutions and reducing the number of parameters to half of that of Faster ARCNN (64k -> 32k)
vinayak19th.github.io
The code repo for this website and instructions for - Creating your own version & Contributing to the Open Source Project
Jekyll BootStrap JQuery Docker SCSS HTML5
Website hosting my resume and profile. Notable Features include:
- System persistent dark mode (Remembers your preference via cookies)
- Live Github based Statistcs
- Profie Statistics on the about page
- Project statistics per repository in the project page
- Reactive design for mobile friendly operation
- Tabbed Page for certifications for easy of reading and management
- Header section with live links such as :
- Direct whatsapp message
- Medium Profile
- Stack Overflow
- etc
- Modular self generating based on yml datasheets
LOW-RESOURCE ASR
A project based on the BABEL Dataset from the US government
Kaldi ESPNet Bash PyTorch
- An End-to-End Framework for performing automatic speech recognition on low-resource languages such as
Urdu and Pashto
- Created the framework based on ESPnet.
BREVIS-NEWS
Open source project to generate a summary of any news article.
Tensorflow Django NLTK
- A Natural Language Processing(NLP) based system to summarize news
articles. The network was trained using ”inshorts” summaries as training data.
- Created a Django based back-end for deployment of the model for edge inference.
Quantum Machine Learning
Implementations of code and projects from the University of Toronto course on Quantum Machine Learning
QisKit Numpy
- Implementing various Quantum Machine Learning algorithms from the University of Toronto Course
- Implemented using the QISKit API from IBM
Neural-Style-Transfer
Neural-Style-Transfer implementation in Tf2 Beta
Tensorflow Numpy Scikit-learn
- Implementing Neural Style Transfer on a Beta Version of Tensorflow 2.0