Computer Vision - Classification

A CT-scan of your CNN

Dog Breed Identification using CNNs

Digits drawn by hand and recognised by using a pre-trained model in android

Image to Text conversion

BERT (Bidirectional Encoder Representations from Transformers) is a transformer-based method of learning language representations. It is a bidirectional transformer pre-trained model developed using a combination of two tasks namely: masked language modeling objective and next sentence prediction on a large corpus.

Image Object Detection using pretrained model FastAPI and UI by Streamlit

Bi-LSTM CRF, CNN, Word2vec, BERT, RoBERTa

Port of the original Pascal VOC 2012 multilabel classification model from caffe to pytorch

Set of VGG neural net models for TensorFlow. Weights converted from Pytorch.

Multi-Task Pre-Training for Plug-and-Play Task-Oriented Dialogue System (ACL 2022)

This is an simple implementation of MobileNet for sign language digit prediction.

Visual Question Answering app with FastAPI and Docker

This is a web applcation built on react and fastapi. It takes an image of a dog as an input and uses deep learning to classify the emotion of dog in three classes (happy, sad, angry)

The Flask-Python web app utilizes a pre-trained image colorization model based on Caffe. It allows users to upload black and white images and applies the colorization model to automatically generate colored versions. The app leverages the power of deep learning to provide an intuitive and interactive way to add color to grayscale images with ease.

Website for the Constellation Dataset project.

Implementation of Finetuning Bert Low-rank Matrix (LoRa) with Gradient Optimizer & Evolution Strategy Optimizer.

Unity project that loads pretrained tensorflow pb model files and use them to predict

This project implements a real-time image and video object detection classifier using pretrained yolov3 models.