Abhijeet Kumar's repositories
Instruction_Finetuning_Tutorials
Train Large Language Models on custom datasets (Tutorials)
Risks_with_LLMs
A collection of risks and mitigation techniques associated with LLM generation.
finbert_embedding
Token and sentence level embeddings from FinBERT model (Finance Domain)
Change-Detection-in-Satellite-Imagery
It employes Principal Component Analysis (PCA) and K-means clustering techniques over difference image to detect changes in multi temporal images satellite imagery.
bert-as-service
Mapping a variable-length sentence to a fixed-length vector using BERT model
bert-embedding
🔡 Token level embeddings from BERT model on mxnet and gluonnlp
Speaker-identification-using-GMMs
It uses GMM to train a speaker identification model. The training and testing has been done on subset (34 speakers) from VoxForge data corpus.
Food-Classification-Task-Transfer-learning
The application demonstrates the advantage of using transfer learning for smaller datasets. Food-11 is a small dataset with 11K images for 11 categories of food. It uses features learnt by already trained VGG16 on large Imagenet data to train a classifier for Food Classification Task..
Web-Crawling-Scrapping-Bot
An application to showcase usage of selenium and beautifulsoup to crawl a website and scrap the desired page in order to extract required infromation
opencv
Open Source Computer Vision Library
QA-system-memory-networks-bAbI-dataset
Building a question and answering system using single layer memory network model on bAbI dataset by facebook. Currently, It is done only for single supporting fact task.
NER_Disease_Extraction_Hackathon
"Innoplexus Online Hiring Hackathon: Saving lives with AI" where disease name has to be extracted from documents. The solution employes a Time distributed BI-LSTM model to perform Named Entity Recognition with 77.8% F1 score.
Predict-the-Happiness-HackerEarth-Challenge
It uses 2-layered fully connected/Dense Neural network model to predict whether the hotel reviews at TripAdvisor site are positive sentiment or negative sentiment.
Toxic_Comment_Classification_Challenge
Building a multi-label classification model which predict a probability for each of the six possible types of comment toxicity (toxic, severe_toxic, obscene, threat, insult, identity_hate)
tf-fashionMNIST
A tensorflow tutorial on Fashion MNIST Dataset. The aim is to showcase an implementation of tensorflow based deep learning model from scratch.
FaceEmotion_ID
Detects Face using Haarcascades and further detects emotion in bounded face (trained a CNN emotion detector model)
Topic-Modelling-on-Wiki-corpus
It uses Latent Dirichlet Allocation algorithm to discover hidden topics from the articles. It is trained on 60,000 articles taken from simple wikipedia english corpus. Finally, It can extract the topic of the given input text article.
Image-compression-with-Kmeans-clustering
A lossy compression to illustrate an application of K-means clustering algorithm
WNS-DS-Hackathon
It's a hackathon problem where we have to predict whether an employee will be promoted or not based on features. (HR Analytics)
Object-recognition-CIFAR-10
It is object recognition task on CIFAR-10 dataset using CNN. It shows how to achieve 90% accuracy.
darknet
Convolutional Neural Networks
Text-classification-and-clustering
It demonstrates the example of text classification and text clustering using K-NN and K-Means models based on tf-idf features.
User-Verification-based-on-Keystroke-Dynamics
To verify a subject's identity based on his keystroke patterns, ie, his typing habits
PyGender-Voice
It uses GMM to train a gender detector model. The testing has been done on subset of Google's AudioSet corpus.
Language-Detection-From-Text---Bi-gram-based
A bi-gram based language model used for language detection task from text
Sentiment-Analysis-using-tf-idf---Polarity-dataset
It uses term frequency and inverse document frequency to do sentiment polarity analysis on movie reviews.
Mail-Spam-Filtering
It uses machine learning models (Multinomial NB & SVM) to predict whether the email is spam or ligitimate on two corpus namely Ling-spam corpus and Euron-spam corpus.