Robotics

Rollie Pollie v1.0

Meet Team-Anything from Department of Mechanical Engineering, BUET, presenting 'Rollie Pollie: A Bio-inspired Rolling-Crawling Robot.' 🤖 Our innovative creation, inspired by the Cebrennus rechenbergi spider, combines walking and rolling for versatile terrain navigation. Enhanced with machine learning, it excels in interactive tasks and surveillance, offering solutions for remote inspections and search and rescue missions. WIth interactive engagements our robot can support you as a pet companion! 🐾

Robotic Arm

Developed a 6-DOF robotic arm controlled via Wi-Fi. Also designed a custom graphical user interface (GUI) with pyQT to control the arm. Integrated real-time control for precise manipulation and efficient operation.

Delta Robot

Developed a delta robot capable of autonomously collecting and sorting different types of plastic bottles (PET, HDPE, and PP) from domestic waste.

Robotic Hand

Developed a flex sensor-controlled prosthetic arm with remote Bluetooth control, enabling precise hand and finger movements. The system provides reliable and responsive operation for assistive applications. Successfully tested the prosthetic arm for real-world functionality.

Phonix v1.0

Phonix v1.0 is a bluetooth control car. I had made it to participate in the "Speed Racing" competition of my university. It is mainly an arduino based project. It's unique chassis design enables it to operate in both normal and upside down position. Speed and Direction of wheel rotation can be controlled by a android app, which is also designed by me.

Mogo v1.0

Mogo v1.0 is an autonomous robo-car. It has two modes: "Obstacle Avoiding" and "Human Following". It uses a sonar sensor while operating in Obstacle Avoiding Mode, and IR sensors while operating in Human Following Mode. It uses an Arduino uno as the main microcontroller.

Line Following Robot

Designed and constructed a line-following robot utilizing the PID (Proportional-Integral-Derivative) control algorithm. The robot autonomously tracks and follows a black path with high precision and smooth navigation.

Soln v1.0

Soln v1.0 is not a real life project. It was a project idea. My team presented it for "R10 Robotics Competition-Ideathon" organized by IEEE. I was the team leader. Our idea was to build a fully autonomous drone that can alleviate human suffering post-calamity situation and forming a bridge between rescue team and survivors. It can also analyze the situation of flood effected area like water level and probability of any epidemic. Our drone will also have some additional use case like monitoring distinct wildlife and reforestation.

IOT

Energy Meter

Designed an Energy Meter using the ESP32. Designed an Energy Meter using the ESP32. This device measures single-phase AC voltage and current and streams the data seamlessly to a webserver for real-time monitoring. This device measures single-phase AC voltage and current and streams the data seamlessly to a webserver for real-time monitoring.

Smart Lock

Developed a "Smart Lock System" with PIN, RFID, and fingerprint unlocking, live camera feed, and intuitive dashboard. Integrated ESP32 for IoT connectivity and added a PIR sensor-based power-saving sleep mode for energy efficiency.

CanSat

Developed a CanSat prototype for atmospheric data acquisition, integrating sensors like temperature, pressure, and air quality. Designed a LoRa-based communication system for long-range duplex communication. Developed a real-time data visualization GUI using PyQt for efficient monitoring and control.

Machine Learning

Pneumonia Detector Website

Created a Website that can predict whether a person is suffering from Pneumonia or not depending on the X-ray image he uploaded. This website works with the help of a custom Machine Learning Model also created by me

Surveillance System for a Parking lot

I made a Image classifier with the help of MLthat can identify whether a parking is occupied or empty. Then I implemented it in the Surveillance System for a Parking lot based on CCTV footage. My model can count the numbers of cars in a huge parking lot and identify which parking are still empty.

Play Flappy bird game with ML

Created a simple "Flappy Bird" game using Python and trained an AI to play it autonomously. Leveraging the NEAT ( Neuro-evolution of augmenting Topologies) algorithm, the AI agent learned to navigate and control the plane through trial and error. After just the 6th generation, the AI became unbeatable at the game, showcasing the incredible power of reinforcement learning algorithms.

Denoising Image with ML

Trained two models using Keras and TensorFlow - one for denoising digit images and another for denoising random images. While the digit denoising model performed exceptionally well, removing noise accurately while preserving image details, the model for random images faced some challenges like decreasing image resolution or smoothing image.

Neural Transfer system with ML

Created a Neural Transfer system leveraging machine learning to analyze and extract painting themes from images, then seamlessly apply these themes to transform other images.

ML Model for Weather Detection

I made a Machine Learning Model that can predict weather from a image (rain/shine/sunrise/cloudy). I used almost 300 sample for each case while training the model. I plotted required graphs to see the performance of my ML and it was satisfactory. Then I tested the model for few random image. The model was working perfectly and can dectect the weather from the images.

Object Detection with ML and Arduinolib

I made a Machine Learning Model that can identify few things like Arduino Uno, Clock, Conch, ESP32, Glue Pot, Highlighter, Notepad, Respberry Pi, Soldering Run and a Toy Umbrella. I tested the model with my smartphone camera as you can see in video. It's performance was not exactly like OpenCv. There is few lacking due to less data point. Then I implemented my model in a arduino library and use it in object detection only using ESP32.

Sign Language Detection

Trained a model for sign alphabet detection (A,L,U,V,W,D,F,7,B,Y) and tested it. It was almost perfect.

Instance Segmentation

Trained a model for image segmentation for a squirrel and tested it. It was almost perfect.

Human Action Detection

Trained a model for human action detection with YOLOv8.

Python

Anisha v1.0

Anisha v1.0 is a "Virtual Assistant". I created it with the help of Python. It can recognize human voices and vocally interact with its user. It can greet you according to date and time. It can play music randomly or according to your choice, send mail, search in Wikipedia, introduce itself, open any application,type on a word file, tell time and date, and provide full weather updates or any individual query for any region of the world. It can be fully controlled with voice commands.

Realtime Facial Tracking

I implemented a face tracking system using MediaPipe, which tracks facial muscles in real-time and plots them as a 3D graph. This technology is ideal for enhancing camera filters and augmented reality applications, efficiently capturing and visualizing facial expressions. The project showcases my ability to integrate and utilize advanced machine learning tools for real-time data processing and visualization.

Face Detection

Done face detection with OpenCV

GraphEye

Developed "GraphEye," a Python-based application for easy data extraction from log-log and semi-log graphs in images. The tool allows users to calibrate for improved accuracy, streamlining the process of reading and collecting data points from graphs.

Object Detection

Done Objeect detection with OpenCV and ESP32cam

Human Tracking

Done Human Tracking from two CCTV footage with OpenCV and data model downloaded from internet.