View on GitHub

The agenda of this project is to apply the improved U-Net architecture on the ISICs data set to predict the segmentation response of the lesion image in the test data set, and the overall minimum Dice score is 0.8.

The data set is challenged by "Analysis of Skin Lesions for Melanoma Detection" using a data set of annotated skin lesion images from ISIC archives. The data set contains raw (input data) and segmented (response data) images of skin patches, which are used to detect skin lesions. The response data includes a binary mask image in PNG format, which indicates the position of the original skin lesion in each input lesion image.

Key Learnings:
• Solving pattern recognition problems via deep CNN
• Deep learning frameworks: TensorFlow and PyTorch
• Algorithmic design with software engineering principles
• Contributing to PatternFlow - a pattern recognition and image processing library
• Mathematical frameworks: group theory and tensor calculus

View on GitHub

GreenMiles is a mobile application designed to incentivize public transportation use while fostering social connections among commuters. The app addresses urban congestion through environmental awareness and gamification. Users can track their carbon footprint through public transport usage, connect with other commuters, earn points for using public transit, and redeem rewards including gift cards and free ride vouchers.

The application was developed using React Native with Expo, TypeScript, MySQL for backend database operations, and Node.js. It demonstrates a comprehensive approach to making sustainable transportation appealing through environmental impact visualization, an incentive structure rewarding eco-friendly commuting choices, and community features fostering peer engagement around sustainability.

Key Learnings:
• Design and build significant software solutions based on design briefs
• Apply multimedia design and software development to deliver polished systems
• Overcome challenges with new designs and technologies
• Apply sound project management skills and tools
• Work successfully in multi-disciplinary teams
• Conduct user investigation, involvement and evaluation
• Apply ethical thinking to technology projects

View on GitHub

LaserTank is an open source computer puzzle game requiring logical thinking to solve a variety of levels, originally developed by Jim Kindley in 1995. In LaserTank, a player controls a tank vehicle which can move forwards, turn clockwise and counterclockwise, and shoot a laser which can be used to interact with special map elements. The game takes place in a 2D map divided into grid cells. The player must navigate from a starting position to the flag in as few moves as possible while avoiding "game over" conditions.

The project involved turning LaserTank into a search problem by defining: (1) a problem state representation (state space), (2) a successor function indicating reachable states (action space and transition function), and (3) a cost function (utility function) with uniform cost of 1 per step.

The solver program takes a testcase filename and output filename as arguments, finds an optimal solution, and writes the solution as a comma-separated list of actions: MOVE FORWARD ('f'), TURN LEFT ('l'), TURN RIGHT ('r'), and SHOOT LASER ('s').

Key Learnings:
• Uniform Cost Search and A* Search algorithms
• Value Iteration (VI) and Policy Iteration (PI)
• Q-learning and SARSA reinforcement learning techniques

View on GitHub

Canadarm2 is a remote-controlled mechanical arm aboard the International Space Station (ISS). The robotic arm is used to deploy, capture and repair satellites, position astronauts, maintain equipment, and move cargo. This project implements motion planning for a simplified 2D version of the Canadarm operating in a planar workspace represented as [0, 1] x [0, 1] C R^2, populated by rectangular obstacles and grapple points which the end effectors can attach to.

Given the initial and goal configurations of the Canadarm robotic arm, as well as a map of the environment, the program finds a valid path satisfying multiple constraints: (1) primitive steps with joint movement limited to 0.001 units, (2) obstacle avoidance, (3) self-collision avoidance, (4) workspace boundaries, (5) joint angle constraints (15-165 degrees), (6) segment length bounds, and (7) collision checking at step boundaries.

Key Learnings:
• Probabilistic Roadmap (PRM) path planning
• Motion planning with multi-joint robotic arms
• Collision detection and constraint satisfaction
• Graph-based search in configuration space

View on GitHub

Developed a complete network protocol suite consisting of three components: (1) RUSHBSvr - a network server capable of sending messages to clients complying with the custom protocol, (2) RUSHBAdapter - works as an adapter for one process through TCP connection, and (3) RUSHBSwitch - functions like a network router supporting both local and global operations.

The final goal for this programming project is sending and receiving data across the global network without losses and errors. Data can be anything attached to the adapter, such as netcat, RUSHBSvr, or even stdin. The implementation demonstrates understanding of Internet architecture, communication protocol design across various layers, network security threats and countermeasures, and RFC standards compliance.

Key Learnings:
• Internet architecture and component functions
• Communication protocol design principles across network layers
• Network security threats and countermeasures
• Computer networking protocol implementation based on RFC standards
• Network traffic analysis using protocol analyzer tools
• Network layers, DNS, HTTP, and file systems

View on GitHub

The team designed and developed an engaging mobile application with minimal cognitive demands for university students that promotes efficient time management by mitigating distraction and aiding task management through incentivisation. The application enables users to study in a more productive manner within the same time constraints.

The functionality of the application is split into four main categories: task management, time management, progress tracking, and incentives. The development followed a rigorous interaction design and human-centered design process completed in three cycles: (1) Contextual Inquiry, Conceptual Design, Low Fidelity Prototype (Paper prototyping) & Initial Evaluation, (2) UX enhancement with Medium Fidelity Prototype developed in Adobe XD & Rigorous Empirical Evaluation, and (3) High Fidelity Prototype (fully functional) with Expert Evaluations from other group members.

Key Learnings:
• Interaction design and human-centered design processes
• Prototyping from low to high fidelity
• User investigation, involvement, and evaluation
• UX design principles and tools (Adobe XD)
• Empirical and expert evaluation methodologies
• Designing for minimal cognitive demands

View on GitHub

A complete simulation consists of three C99 programs: (1) roc2310 - simulates an aircraft, (2) control2310 - simulates an airport, and (3) mapper2310 - maps airport names to ports. Multiple instances of these programs are executed concurrently to carry out the simulation. The project extensively used pthreads for concurrent programming and communicated via IPv4 TCP networking with blocking communications.

Key Learnings:
• C programming compilation and initial features
• Process management and inter-process communication
• Virtual memory concepts
• File descriptors and pipes
• Thread programming and synchronization techniques
• Network layers and protocols
• DNS and HTTP protocols
• File systems architecture

View on GitHub

The program is a simple version of the classic Asteroids game. The AVR ATmega324A microcontroller runs the program and receives input from a number of sources while outputting a display to an LED display board. Additional information is output to a serial terminal, and as part of this project implementation, a seven-segment display and other LEDs were integrated to enhance the gaming experience.

Key Learnings:
• AVR microcontroller programming
• LED matrix display control
• Hardware interfacing and I/O operations
• Serial communication protocols
• Real-time embedded systems programming
• Seven-segment display implementation

View on GitHub

The program models a public transportation network with a comprehensive class hierarchy. Main classes include: (1) Passenger - a standard passenger with ConcessionPassenger subclass, (2) PublicTransport - defines basic vehicle functionality with Bus, Train, and Ferry subclasses, (3) Stop - responsible for routing passengers to their destinations, and (4) Route - defines basic route functionality with BusRoute, TrainRoute, and FerryRoute subclasses.

Key Learnings:
• Object-oriented design principles
• Class hierarchy and inheritance
• Network modeling and graph structures
• Java programming best practices
• Transportation systems analysis

View on GitHub

A desktop application written entirely in Python using Tkinter for GUI development. When a user presses either the "Request Quick Help" or "Request Long Help" button, the application displays a dialog box that asks the user for their name. Once entered, the application joins the user to the queue, displays their name on screen, and regularly updates the amount of time they have been waiting (roughly once every ten seconds). The "Questions Asked" heading tracks questions answered since opening the application, maintained separately for quick and long help categories.

Key Learnings:
• Python GUI development with Tkinter
• Queue data structure implementation
• Real-time UI updates and event handling
• Dialog box implementation
• Time tracking and statistics management