Shubhra Deshpande's repositories
libraryForStereoVision
Recreated the OpenCV library function for stereo vision, Homography matrix and segmentation from scratch using python
TomJerryGame
Tom and Jerry game is based on reinforcement learning. Tom is supposed to learn the best and efficient most path to reach Jerry in the 5 x 5 mesh by learning every possible path. Developed a game in python where Tom cognitively learns the best strategy to reach Jerry using Deep Q-network (DQN) of R.L. The agent uses Q-table to understand the rewards or punishments.
alexa_base_skill
Base code for developing an Alexa skill.
alfcapp
App for Aurangabad Football Network
Autotranscription
Designed the Wireframes (Lo-fi) sketches on Autodesk Sketchbook. Created the Wireframes (Hi-fi) on Adobe XD. The prototyping is also done in Adobe XD.
aws_rekognition_flutter
The python script runs over aws lambda, being powered by serverless. AWS image Rekognition microservice supports face comparision and the json file is read in flutter app as deployment
aws-cloudformation-resource-providers-ecr
The CloudFormation Resource Provider Package For Amazon ECR
documentation
The source for Datadog's documentation site.
draggable_widget_demo_video
Flutter app demo
edge_detection
We see that image is a two dimensional array of pixels. The task of edge detection is performed using sobel operator. Operating sobel operator on image and collecting the edges in ‘x’ and ‘y’ direction, program calculates the norm i.e. root mean square distance of every pixel of both matrices and define the resulting pixel as edge defining agent. Very crude method of logical hard coding is used to complete this task hence no external library has been used. The task can be seen as collection of three main functions -Defining kernels or sobel filters for X and Y direction -Defining convolution function -Calculating norm Defining Kernels: Functions kernel-x(), kernel_y() define sobel operators as shown in figure 1.a the sobel operator Kernel Gx filters out all the Y leading points and hence giving horizontal subpart of the image Kernel Gy filters out all X leading points and gives vertical defining edges
flutter_python_poc
This app is a small demo about wiring up python files with dart specifically flutter app and flash the output
HandwrittenDigitRecognizer
Continued the project of UB of detecting and recognizing the handwritten digits in the address space over the letterhead in USPS data. - The convoluted neural network (CNN) was implemented with tensorflow in python. - The accuracy of the model was increased to 49%
houghTransformUsingPython
We have a predefined library for implementing Hough transform in OpnCV. But this project delivers the developed library function for edge and circle detection using Hough transform algorithm in python.
kmean_clustering_baboon
K mean is implemented using dictionaries. The project contains two code files. The first code, k_points.py, is for the simple k mean clustering applied on few random points. The second code file named k_baboon.py contains k mean clustering code for clustering the famous baboon image into 3,4 and 5 clusters respectively.
move_child_to_room_alexa-skill
move child/ children from one room to other room
personalized_video_suggestion
Personalized Video Summarization. Where given an android app, the user can watch selected youtube videos and then group them into different clusters (just like bookmarking). These user preferences are learnt by the clustering algorithm to adjust the boundaries of clusters so that better video suggestions could be given to the user.
SurveyGuide
App created in flutter that lets you take small surveys
UniqueAuthorRecognition
The author of the handwritten sample was predicted by handwriting analysis using CNN with Keras. The project is also implemented with logistic regression and linear regression using nonlinear gaussian basis functions.
Validity_challenge
web page displaying the duplicates in the csv records
youtube_8m_video-id
get youtube video links from youtube 8M tfrecord files