Research Paper Classification Using GCNs
- Implemented Graph Convolution Networks (GCNs) on the Cora dataset, a graph dataset consisting of interconnected scholarly articles via links, to classify scientific papers, boosting classification accuracy from 73.16% to 83.52%.
- Analyzed how GCNs utilize the relationships between articles to enhance the accuracy of scientific paper classification.
SilkViser: Visualizing Blockchain Cryptocurrency Data
- Designed SilkViser, a dashboard facilitating the exploration of Bitcoin cryptocurrency data through visualizations like coin glyphs, sankey charts, and choropleth maps, enhancing comprehension of transaction mechanisms and network relay activities.
- Utilized HTML, CSS, JavaScript, and d3.js to build SilkViser, incorporating real-time API calls for up-to-date transaction data.
Student Verification System
- Developed an e-verification portal for the college which consists of an AI based attendance system and a verification panel.
- Attendance system stores the attendance of the students in an excel sheet using face detection and recognition.
- Verification panel helps in verifying and filtering students based on the criteria setup by companies participating in college placement drives.
Face Recognition System
- Studied and implemented Face Detection using Haar Cascade Classifier and Face Recognition using Eigenfaces and Invariant Scattering Convolutional Networks
- Compared the results obtained by using Eigenfaces and the Scattering Networks approach and generated the ROC and CMC curves with 94% and 95% respectively
- Worked on the LFW and Yale face dataset for the above implementations
Result Processing System
- Developed a Result Processing System website for the college in PHP that calculates and provides the result of a student semester wise in marksheet format.
- Further connected the website with a database in Xampp server that stores the complete information of the student along with course details.
Minutiae Extarction from Distorted Fingerprint Image
- Studied about fingerprint classification and implemented it using python codes
- Enhanced the fingerprint image using Gabor Filter, then applied thinning on the enhanced image and extracted minutiae (ridge bifurcation and ending points) from the thinned image for classification