supekar Goto Github PK

configcli

driver-inattentiveness-detection-system-using-opencv

The driver inattentiveness detection system on raspberry pi is implemented by using the haar cascade classifier in OpenCV. The facial features are detected using the facial landmark detector of dlib library.

emotion-detection-on-raspberry-pi

Timed.py takes in images fed to it from raspberry pi camera module. It then extracts the emotions detected in the faces. The information regarding the the number of people and their corresponding emotions is assembled into a JSON packet as per the protocol set between the client side and server and sends the JSON packet to the server. This occurs periodically every 60 seconds according to the settings of the scheduler in timed.py

hidden-markov-model

This is an implementation of first order Hidden Markov Model for speech tagging. Feature engineering is done to create the 3 matrices for HMM model training, prior, emission and transmission matrices. These matrices are then used by forwardbackward.py that implements a first order HMM by using the forward backward algorithm. The program proceeds to calculate negative log likelihood and accuracy and stores them in a file

kubedirector

Kubernetes Director (aka KubeDirector) for deploying and managing stateful applications on Kubernetes

logistic-regression

Logistic regression on movie review polarity dataset. feature_engineering.py implements feature engineering on the existing data set to convert it into a format that is suitable for processing to perform logistic regression. The logistic regression takes the feature engineered data of train, test and validation data to perform binary logistic regression on it. The accuracy and negative log likelihood were written into an output file

low-power-audio-surveillance-system

mkdocs-bootstrap

MkDocs Bootstrap Theme

neural-net

Implemented a 1 hidden layer neural net with variable number of epochs, learning rate and number of hidden layer with numpy. neural.py implements the back propagation algorithm to update the weights. A final pass on the testing data after trainings the weights on number of epochs gives us the accuracy of the model on classification of 10 output labels.

test

test repo