Souvik Datta's repositories
Book-Bot
This is a robot with Barcode + QR Code scanning capabilities designed for collecting books and navigation across a campus. The robot will mould the use of Computer Vision for ensuring safe collection and identification of books, all this while being powered by a Raspberry Pi for faster processing.
Semantic-Segmentation-of-Satellite-Imagery
• Devised a U-Net-based Image Segmentation Model for land cover analysis on 5 classes from LandCover Dataset. • Developed a network based on the ResNet34 Architecture and achieved 0.72 mean Intersection over Union (IoU).
Soil-Type-Classification
• Designed a classification model with 5 classes and 98% testing accuracy using a Convolutional Neural Network. • Applied Data Augmentation and reduced validation losses by 30% by applying MobileNetV2 Architecture.
Traffic-Sign-Classifier
An autonomous braking system that uses sensor data from Radar and Computer Vision to localize the accurate position of an object around a vehicle using Neural Networks. An algorithm to fuse the data from sensors and camera to perform odometry operations
3D-Point-Cloud-Map
• Produced 3D point cloud data maps from 2D monocular RGB image-based depth maps using Open3D functions. • Exploited a PointNet based Object Detection model to classify 3D point cloud data with 81% training accuracy.
Jet-Engine-Health-Prediction
• Predicted Remaining Useful Life (RUL) for turbofan jet engines using a Random Forest Classifier Model. • Performed Exploratory Data Analysis on NASA’s Dataset and analyzed feature correlation using Seaborn. • Applied hyperparameter optimization using Randomized Search CV and increased F1-Score to an overall 91.2%.
Grounded-Segment-Anything
Marrying Grounding DINO with Segment Anything & Stable Diffusion & Tag2Text & BLIP & Whisper & ChatBot - Automatically Detect , Segment and Generate Anything with Image, Text, and Audio Inputs
MLProjectTraining
Collecting dataset for training my model for recognizing and Identifying Various Objects
Reinforcement-Learning-Robot
• Programmed a quadruped robot to exhibit walking gait using PyTorch-based Reinforcement Learning model. • Optimized learning rate and other hyperparameters to marginally enhance Mean Reward of the model by 14%.
Kalman-Filter
• Developed Kalman Filter-based state estimation model for precise approximation & visualization of a robot’s state. • Attained an extremely low Mean Square Error of 2.8% & 1.1% for the robot’s position & velocity on evaluation.
Awesome-Implicit-NeRF-Robotics
A comprehensive list of Implicit Representations and NeRF papers relating to Robotics/RL domain, including papers, codes, and related websites
bark
🔊 Text-Prompted Generative Audio Model
Image-processing-for-Hacktoberfest
Uploading images for collection
Open_for_Contributors
Opportunity to start open source and PR experience
OpenSourceEnthusiast
Showcase Your Programming Skills here without hesitation
python_codes
This is a beginner friendly , repository made for python codes, specifically for Hacktoberfest event
whisperX
WhisperX: Automatic Speech Recognition with Word-level Timestamps (& Diarization)