Sai Muralidhar Jayanthi's repositories
time_series_classification_prediction
Different deep learning architectures are implemented for time series classification and prediction purposes.
CodemixedNLP
CodemixedNLP: An Extensible and Open NLP Toolkit for Code-Switching
object_tracking_using_pan_tilt_platform
Target tracking is the process of locating a moving object throughout a sequence of frames in a video. The project aims to design and realize a tracking system using a camera mounted on a controlled pan-tilt platform. As the target moves in the frames of the camera, the pan-tilt platform aligns itself in such a way that the object being tracked remains at the center of the camera's frame and thus in its field of vision. Clustering of Static-Adaptive Correspondences for Deformable Object Tracking (CMT) is an award-winning object tracking algorithm which is able to track a wide variety of object classes in a multitude of scenes without the need of adapting the algo- rithm to the concrete scenario in any way. A C++ implementation (CppMT) is freely available under the BSD license, meaning that you can basically do with the code whatever you want. References [1] Nebehay, G., Pugfelder, R.: Clustering of Static-Adaptive correspondences for deformable object tracking. In: Computer Vision and Pattern Recognition, IEEE (2015) 2784-2791
object_tracking_under_occlusion
Skin-colored objects are detected with a Bayesian classifier which is bootstrapped with a small set of training data. Then, an off-line iterative training procedure is employed to refine the classifier using additional training images. On-line adaptation of skin-color probabilities is used to enable the classifier to cope with illumination changes. Tracking over time is realized through a novel technique which can handle multiple skin-colored objects. Such objects may move in complex trajectories and occlude each other in the field of view of a possibly moving camera. Moreover, the number of tracked objects may vary in time. A prototype implementation of the developed system operates on 720x1080 pixel video with a frame rate of 30 per second.
hdlss_techniques_classification
The Gene Expression Omnibus (GEO) data series GSE4115 contains data from 192 human subjects, each with 22,283 profiled genes. Each subject can have one of three disease states: cancer, no cancer, or suspected cancer. Your task is to build a classifier for cancer vs. no cancer by using HDLSS techniques (such as elastic net).
constrained-fact-verification
EMNLP 2020 paper (Constrained Fact Verification for FEVER)
morpheus_multilingual
SIGMORPHON 2021 Paper (A Study of Morphological Robustness of Neural Machine Translation)
11-767
Course materials for 11-767
digital_image_processing
DIP Techniques: 1.Edge preserving smoothing Filters 2.Image Restoration 3.Unitary Transform
eacl2021-OffensEval-Dravidian
EACL 2021 paper (SJ_AJ@DravidianLangTech-EACL2021: Task-Adaptive Pre-Training of Multilingual BERT models for Offensive Language Identification)
GLUECoS
A benchmark for code-switched NLP, ACL 2020
jekyll-now
Build a Jekyll blog in minutes, without touching the command line.
keyword_extraction
An NLP algorithm to extract keywords from a given piece of plot/ description/ summary.
lstm_text_prediction
Train an LSTM on Human Action by Ludwig von Mises . This book is supposedly the best defense of capitalism ever written. Then generate five samples of random text that sound like his work. This code works with a fair understanding of LSTM and its implementation using keras library with theano backend. It is recommended to run the python script using GPU, as recurrent models are quite computationally intensive.
memory_management_techniques
CPP code for FIFO, LRU and NFU memory management techniques. 'main.cpp' contains 2 global variables viz. 'n' and 'MAX_RANGE' which define the input number of queries and max_range of each query respectively.
movie_recommendation_system
A recommendation engine built using movies-lens 100k dataset. Given a user id, the system obtains movie features for all the movies in the dataset (from the already-trained-and-saved latent features obtained from user-item watchlist and optimizing a loss function by back propogation) and finds a list of best movies in the interest of the user based on his watch-history using a cosine similarity.
murali1996.github.io
A beautiful, simple, clean, and responsive Jekyll theme for academics
neural_nets_with_numpy
Modular implementation of forward and backward propagation in multi-layer-perceptrons and convolutional-neural-networks with different loss functions and activation functions
semantic_segmentation_of_nuclei_images
UNET architecture implementation with data augmentation techniques