Bengaluru, Karnataka, India
5K followers
500+ connections
About
Experience & Education
View Paras’s full experience
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
Projects
-
Multi Class Text Classification
Categorization of companies based on their domain and functionality
Involved all steps from Data pre-processing - text cleaning , feature engineering to model building and then evaluating the model using metrics - accuracy, recall and precision
Experimented with different features such as Count Vectors, TF-IDF Vectors, Word Embeddings
Trained and evaluated different classifiers - SVM, CNN, LSTM using the above features -
Time Series Forecasting using Deep Learning
Demonstrated minimum 15% increase in accuracy of E2open Forecasting Solutions using LSTM
Involved use of tools, like graphing and summary statistics, to better understand the data
Review plots and summarize and note obvious temporal structures, like trends seasonality, anomalies like missing data and outliers
Use of techniques like ADF test, ACF and PACF to check stationarity of Time Series
Secured first rank among 46 teams from different countries -
Liveness Detection for Face Recognition
Analyzed different techniques for liveness detection like frequency and texture based analyses, blinking based analysis
Implemented texture based analysis using Local Binary Pattern (LBP) for discriminating 2-D paper masks from live faces in MATLAB
Implemented novel hybrid approach in which eyes are detected in sequential input images taken from a video database and cropped thereafter, and texture information from each frame is taken using LBP, Local Ternary Pattern (LTP) and 3D LTP…Analyzed different techniques for liveness detection like frequency and texture based analyses, blinking based analysis
Implemented texture based analysis using Local Binary Pattern (LBP) for discriminating 2-D paper masks from live faces in MATLAB
Implemented novel hybrid approach in which eyes are detected in sequential input images taken from a video database and cropped thereafter, and texture information from each frame is taken using LBP, Local Ternary Pattern (LTP) and 3D LTP technique
Used SVM for differentiating the feature space into live and non-living -
Personal Authentication using Gait Recognition
Model Free Biometric Recognition
Feature Engineering on Silhouette Images from CASIA Database
Extracted different features using methods such as Shannon Entropy, Pal and Pal Entropy, Centroid Method, Active Energy Image
Applied techniques like PCA for data compression, LDA and SVM for classification
Other similar profiles
Explore more posts
Explore collaborative articles
We’re unlocking community knowledge in a new way. Experts add insights directly into each article, started with the help of AI.
Explore MoreOthers named Paras Ahuja in India
95 others named Paras Ahuja in India are on LinkedIn
See others named Paras Ahuja