Saumya Bhandary's repositories
Naive_Bayes_Classification
This classifies the given statement into which categories they belong to.
ThePathfinder
A Computer Vision Based Navigation System which uses Differentiable Feature Clustering to Segment the road. And uses different ROI based methods to figure out road, and detect obstacles in it and provides any kind of self driving system to either go ahead or stop.
Saumya-Bhandari
A portfolio for representing myself to the world.
ANN_NumericDigitClassification
A neural network made and analyzed, with 1 hidden layer, 2 hidden layers and 5 hidden layers and was trained and tested on MNIST dataset. The Initializaton method used was Xavier, the PCA was done using SVD and 3 proper neural networks were made.
ID3-From-Scratch
ID3 algorithm, which uses entropy and Information gain was created on the samplecar.csv dataset.
Variational-Auto-Encoders
A small implementation of variational autoencoder setup, in convolutional neural networks. I tried to learn generative modeling, and in that used this setup to train a neural network to restore given images.
Image-SuperResolution-
A Super Sampling model created using the SRCNN method proposed by Chao Dong, Chen Change Loy in 2015. It uses Convolutional Networks to identify features and uses "Depth-To-Feature" technique in the end to generate a high resolution image of a given low resolution input. The model is trained and tested on BSDS500 dataset.
Initialization_Techniques_and_Other_Findamentals_of_Neural_Network
This notebook consists of initialization techniques and other fundamentals of Neural Networks.
Moore_Penrose_PseudoInverse
The notebook here consists of the concept of "Moore Penrose PseudoInverse" which approximates the inverse of a matrix and we use it to find the weight vectors in any given linear regression problem.
Music_Genre_Prediction_Using_Different_ML_Models
We use different models to predict the type/genre of music given to us.
Principal_Component_Analysis
This repository consists of performing Principal Component Analysis (Finding out the principle components of our dataset), using the concept of Eigen Vectors and Eigen Values of the correlation matrix. Also, it also includes the technique of reducing the dimension of the data using Singular Value Decomposition technique and reduce it to the lowest possible dimension.
PyTorch_Boilerplate_Train_Test_Valid
This repository consists of a boilerplate code for initializing model, training the model, testing it and also boilerplate for the dataloader class which returns the train, test and valid data.
Image-Segmentation-Network-Using-CNN-in-a-UNet-Architecture
A simple segmentation network, built using PyTorch, which is easy to use for your basic segmentation task. This project was made to segment the road-area from any given image to implement it in self-driving systems.
Support_Vector_Machine
A SVM classifier created to classify data on the IRIS dataset. A linear SVM as well as a Radial Basis SVM was created and also a polynomial kernel was created which classifies the data correctly to their original class.
Convolution_Neural_Network_For_Digit_Classification
Built a CNN for classifying handwritten digits using PyTorch and used the GPU to optimize the network to boost up the performance. A visual representation of the CNN model is also included in the notebook.
Game-Of-Life-Simulator
Created a small simulation of Conway's "Game of Life", just for fun.
Image-Classification-Network
Created a 16 layer Convolutional Neural Network model with 3 connected layers from scratch in CUDA to determine whether the image given is an image of a Sandwich or a Pizza.
Image_Processing_With_Convolution_Operation
Image Processing done analyzing different filters, with different properties, to blur, detect edges, emboss and perform other image processing filters.
mandlebrot
Trying out to Render Mandlebrot set around.
medical_mae
The official implementation of "Delving into Masked Autoencoders for Multi-Label Thorax Disease Classification"
MiraSite
A website developed for a friend to post her poems and stories. Kind of like a personal blog.
Optimization_CUDA_From_Scratch
Created a optimization concept with Mini-Batch Gradient Descent and Used Ada-Grad as an optimizer and implemented the entire training process in CUDA for performing the training of the an ANN with 4 hidden layers.
PlantDisease_Knowledge
A Starting knowledge base on Plant Diseases
XrayGPT
XrayGPT: Chest Radiographs Summarization using Medical Vision-Language Models.