Projects

Course projects:

• Pedestrian Detection in Thermal Images, December 2017 - January 2018
  • As part of 'Neural Networks and Fuzzy Logic' course under Dr Surekha Bhanot
  • Implemented an adaptive fuzzy clustering and CNN based model to detect bounding boxes around pedestrians from LSIFIR dataset.
  • Upon clustering using adaptive FCM, the image pixels were binarized and the contours in the image were detected to draw candidate bounding boxes. The candidates were pruned and cropped out of the image and classified as a hit or a miss using the CNN which detected presence of human in the crop image.
  • Using PASCAL measure, the miss rate achieved was 3%.
• K - Nearest Neighbour Clustering in Scheme, October 2017 - November 2017
  • As part of 'Principles of Programming Languages' course under Dr. Lavika Goel
  • Implemented KNN clustering in scheme using euclidean distance measure. Query system could also be used to extract noise, border and core points in the system.
• Pose Aware Face Recognition in Wild, January 2017 - May 2017
  • As part of 'Machine Learning' course under Dr. Kamlesh Tiwari.
  • Worked on "Labelled faces in Wild" dataset to detect and recognise faces from non frontal images in the wild.
  • Developed a model to load images as NumPy arrays and generate 'eigen-faces' using PCA and SVM. This model achieved accuracy of 91% on the dataset. [GitHub]
  • Also learned to implement CNNs in TensorFlow to achieve accuracy of 99%. This model is described here.
• Smart AC System, January 2017 - May 2017
  • As part of 'Microprocessors and Interfacing' course under Dr. Nitin Chaturvedi.
  • Designed and developed a prototype temperature monitoring and AC control system using Intel 8086 microprocessor.
  • Detailed report : Report
• Query Answering System in Prolog, October 2016 - November 2016
  • As part of 'Formal Logic in Computer Science Course' under Prof. Sundar Balasubramaniam.
  • Developed a self answering system in Prolog, which could answer questions related to a particular section of BITS academic guidelines.
• Automated Algebraic Expression Simplifier in Prolog , November 2016 - December 2016
  • As part of 'Formal Logic in Computer Science Course' under Prof. Sundar Balasubramaniam.
  • Prolog based Algebraic Expression Simplifier based on basic identites. Motivation gained from ABSolver.
  • The input was multivariate expression in standard Reverse Polish notation. It was parsed into the code which applied various algebraic identities and returned the expression as sum of powers of variables. [GitHub]

Study projects:

• Automated Crack Detection Civil Structures, August 2018 - Ongoing
  • Ongoing design project under Dr. Dhiraj Sangwan At CEERI, Pilani
  • Aim to develop a quasi real time automated crack and defect detection system for civil structures
• Trust Management in Wireless Sensor Networks, December 2016 - May 2017
  • A semester research/study project under Professor Ashutosh Bhatia at BITS Pilani to analyse the currently deployed trust management schemes in Ad-Hoc and P2P networks and test their efficacy for deployment in WSNs.
  • Simulated Wormhole, HELLO flooding and selective forwarding attacks using “Low Energy Adaptive Clustering Hierarchy” topology for the sensor network on Castalia simulator based on OmNet++ platform.
  • Detailed report : Report

Other projects:

• Freudian Slips, January 2017 - February 2017
  • A Universal Windows app for cognitive growth testing in subjects which can not be tested using regular means.
  • Exposed subject to an 'affective' image and capture facial reaction as an image which was used to analyse cognitive reaction to the image and matched with GAPED dataset.
  • A self devised mathematical model was used to judge abnormalities in reaction. The final result contained suggestive actions for the diagnosed problems. [GitHub]

• Automated Horn Regulation Mechanism, December 2015 - February 2016
  • Fabricated a device to be fitted into automobiles for automated horn regulation to prevent unnecessary use of horn.
  • The device used a microcontroller fetching gps location in real time and mapping that across various noise sensitive zones in the city. If the vehicle breached a certain maximum distance barrier, the horn volume was reduced to zero. The device incorporated mechanisms for emergency where the use is required.
  • The project won first prize at APOGEE 2016- BITS Pilani’s technical fest and took part in finals of Smart India Hackathon '17. [GitHub]
• Weekly Progress Report Management Portal, May 2017 - July 2017
  • As part of summer intern at IDS Infotech, Mohali.
  • Developed a web portal for automated filling and viewing of weekly progress reports which were used by company for tracking project progress.
  • Detailed reports : Report

For code and implementation details of any of the following projects, check out the links in the description or get in touch with me. I shall help you out.