s3pi Goto Github PK

alliance_univ_workshop_2

Sementation and GenAI

bayesian-classifier

Bayesian Classifier for 3 types of data on 4 different scenarios of Covariance matrices. Data types: i) Linearly Separable Data, ii) Non Linearly Separable Data and iii) Real World Data. 4 scenarios are i) Covariance matrices of all 3 matrices are same and equal to kxI ii) Full Covariance matrices are all 3 classes are equal. iii) Covariance matrix is diagonal and different for each class. iv) Covariance matrix is full and different for each class.

building-gpt

Lets build a model from scratch into a Transformer architecture.

clustering-kmeans-and-gmm

Clustering using KMeans and GMM algorithms.

computer-vision

Computer Vision assignments on Image Transformation, Estimation of Transformation Matrix, Estimation of Fundamental Matrix and Estimation of blur from depth.

digital-image-processing

Digital Image Processing assignments on Image Corrections, Filters, ICM and others.

dtw-hmm

Template matching for variable sequence length using Dynamic Time warping and Hidden Markov Model algorithm.

enhancing-lowdose-ct-scan

Denoising Lowdose CT Scan

finetune-densenet121-on-nih-xray-data

Finetune Densenet_121 trained on Imagenet on public XRay data available at NIH. This is expected to help with better feature extraction for XRay classification.

football-game-prediction

Predict the outcome of a football match

gpt_from_scratch

iris-segmentation-and-model-interpretation

Pixel wise segmentation of Iris using Encoder-Decoder architecture and the interpretation of the model's decision using Layer-wise Relevance Propagation (LRP) algorithm.

k-nearest-neighbours

K-Nearest Neighbor( kNN) is a non- parametric algorithm which is used for classification as well as density estimation.

masters-thesis

Medical Imaging techniques for Transformation and Inference using Deep Learning

ml_flow_end_to_end

Deploy ML model end to end

ml_flow_end_to_end_part_1

mr-image-reconstruction

Semantic features aided multi scale reconstruction of inter modality MR Images.

perceptron-fda-svm

Comparision of classification performance on data projected using FDA, Perceptron and SVM algorithms.

principal-component-analysis

Principal Component Analysis (PCA) technique is used to reduce the dimension of the data while keeping the information of the original features as close as possible.

s3pi

shape-from-defocus

Reconstruct an all-focused image from several de-focused images.

swinbert

Research code for CVPR 2022 paper "SwinBERT: End-to-End Transformers with Sparse Attention for Video Captioning"

xray-to-report

Hierarchical X-Ray Report Generation via Pathology tags and Multi Head Attention