Machine learning, Python.
Resume:
Sameeksha M Vernekar
******************@*****.***
CAREER OBJECTIVES
A firm believer in hard work, perseverance and an ardent learner, eager to work with new and different technologies. Aspire to gain a hands-on experience and industry exposure by working on a live project and use acquired skillset to contribute on solving real world issues.
ACADEMIC QUALIFICATION
Year Examination Institute Board/
University
Percentage
2017-21
Bachelor of
Engineering
(Electronics and
communication)
KLE Technological
University, Hubli
Deemed
University
7.06 CGPA
2017 XII (Science) Impulse Pre-
University
College, Hubli
Karnataka 68.16 %
2015 X (General) St. Antony's Public
School, Hubli
CBSE 9 CGPA
TRAININGS AND WORKSHOPS
2020 - Introduction to IOT & Embedded Systems
2020 – TCP/IP and Advanced Topics
2020 – Machine learning with TensorFlow on Google Cloud Platform Specialization KEY PROJECTS
Deep Learning Framework for Detection and Localization of Object Towards Land Survey: The aim of the project is to generate Data set for detection and localization of object towards land survey, semantic segmentation of specified object towards inference. We have developed a pipeline to detect and localize the objects. Towards this, we aim to address a fundamental yet challenging problem on the detection and localization of the specified objects in an drone captured images. To achieve this, first we implemented object detection algorithms like faster-RCNN, YOLO v3, and further we implemented semantic segmentation algorithm SegNet. The data set we collected from InDrone start-up which are drone captured images. Towards this, we feed drone captured images for different architectures to detect the objects and localize the objects in drone captured images. The drone captured images are captured at a height of 120 meters, the drone captured images have the resolution of 4864 X 3648. These captured images are annotated for object detection and semantic segmentation algorithm using available GUI tools. To detect coconut trees, we implemented object detection algorithms like Faster-RCNN and YOLOv3, also we performed crop at test for Faster-RCNN algorithm which gave good results compared to YOLOv3. To localize the objects, we implemented semantic segmentation algorithm SegNet. Our object detection algorithm gives more precision compared to EAVISE Research Paper 2017 which shows that 0.71 mAP, and we achieved a land survey by implementing SegNet algorithm with 0.75 precision. Model encryption through Data Obfuscation:
The aim of the project is to ensure the privacy, authenticity and certainty of data, to prevent information from manipulation. The objective was, 1. preserving the data of proposed model from pirate users. 2. Predicting the health status of a person. 3. Encrypting the data through data obfuscation. 4. Maintaining the accuracy of the model. The model uses two parameters i.e., height and weight. Based on these parameters the model predicts if a person is healthy or unhealthy. It uses logistic regression for the prediction. A training dataset is necessary to train and generate a machine learning model. The prediction of the health status of the test data depends on the trained data. For the security issues of our model, it is important to encrypt the data before submitting them to the customers. Henceforth, we propose RSA algorithm. RSA is an algorithm used to encrypt and decrypt messages by modern computers. It is an algorithm for asymmetric cryptography. Asymmetric means two different keys are available. This is also called public key cryptography, because anybody can obtain one of the keys. It is necessary to keep the other key confidential. The algorithm is based on the fact that it is difficult to now the factors of a large number of composites: the problem is called prime factorization when the integers are prime numbers. It is also a generator of key pairs
(public and private key). The accuracy obtained is 82.44%. Learning based estimation of attenuation coefficients towards underwater image restoration: This is an on-going project.
Here we work on the objective to do literature survey towards variants of image formation models in underwater scenario, techniques to identify the attenuation coefficient towards modelling restoration, experimenting with benchmark datasets towards demonstrating results of restoration with estimated attenuation coefficients. The attenuation coefficient is finding atmospheric light, the depth map, and the transmission map of an underwater image. Using these attenuation coefficients, we perform restoration of an image.
TECHNICAL KNOWLEDGE
SKILL SET
• Good knowledge on C programming, Python language and Linux OS.
• Familiar with practical knowledge on image processing technique/machine learning/computer vision/deep learning.
• Hands-on experience on coding ML models using TensorFlow/OpenCV platform.
• Good knowledge on working with Embedded System Design using Raspberry-pi module.
• Good communication skills in both written and verbal.
• Work successfully in a team environment as well as independently. OTHER INFORMATION
Father’s name Maruti M Vernekar
Mother’s name Shobha M Vernekar
Date of Birth 25 May 1999
Languages Known English, Kannada, Hindi, Konkani
Address #18, Daivajyna colony, Siddeshwar park, Vidyanagar. Hubli - 580031 Declaration: I hereby declare that the above provided information is true to the best of my knowledge. Date: September 25, 2020
Place: Hubli
Contact this candidate