← antimony labs

Udacity portfolio

Coursework projects from across four Udacity nanodegrees (Self-Driving Car, Robotics, C++, Deep Learning) plus a few ML/CV one-offs. Each writeup covers what was built, what worked, what didn't, and what I'd change with the benefit of years of hindsight.

21 projects · sources archived on GitHub

Self-Driving Car Nanodegree · 5

  1. computer vision
    Lane Finding (Project 1)
    The first SDC nanodegree project — Hough-transform lane detection on highway video with classical OpenCV. Gateway drug to computer vision.
  2. slam localization
    Extended Kalman Filter
    Fuse LIDAR + RADAR with an EKF to track a moving vehicle. C++, uWebSockets, Udacity simulator. The first project where the math actually mattered.
  3. computer vision
    LIDAR Object Clustering
    Filter ground, build a k-d tree, Euclidean-cluster obstacles. C++ from scratch on PCL — no ML, all classical geometry.
  4. slam localization
    Particle-Filter Relocalization
    Markov localization via particle filter — given a known map and noisy LIDAR observations, estimate the vehicle pose. The first time Bayes felt physical.
  5. deep learning
    Traffic Sign Detector (TensorFlow)
    CNN classifier on the German Traffic Sign Recognition Benchmark — the Udacity SDC project that introduced TensorFlow before Keras was the default.

Robotics Nanodegree · 3

  1. robotics ros
    PR2 Perception Pick-and-Place
    ROS + PCL stack for the PR2 robot: cluster a tabletop, classify objects with an SVM on color/normals features, generate grasp poses.
  2. control planning
    PR2 Kinematics Pick-and-Place
    Solve the PR2's 7-DOF inverse kinematics symbolically, then drive arm-joint trajectories to pick and place objects. Sympy + ROS + tears.
  3. robotics ros
    Swarm Robotics Project
    Multi-robot coordination in ROS — N TurtleBots converging to a formation pattern via decentralized reactive control.

C++ Nanodegree · 5

  1. control planning
    OpenStreetMap Route Planning (A*)
    Hand-rolled A* over real OpenStreetMap road graphs in C++. The capstone of the C++ nanodegree's foundations module.
  2. systems cpp
    Concurrent Traffic Light Simulator
    Multi-threaded traffic-intersection simulator in C++17 — the C++ nanodegree's capstone on concurrency, futures, and message queues.
  3. systems cpp
    Linux System Monitor (htop clone)
    Parse /proc to build an htop-style live process viewer in C++17 with ncurses. The C++ nanodegree's intro to Linux internals.
  4. systems cpp
    Snake (SDL2 game)
    Capstone of the C++ nanodegree foundations module — the classic Snake game in SDL2 + modern C++ with score persistence.
  5. systems cpp
    Memory-Management Chatbot
    Refactor a leaky C++ chatbot to demonstrate every smart-pointer + ownership pattern in the language. The capstone of the C++ memory module.

ML / CV (other coursework) · 8

  1. computer vision
    pyAprilTag (Python wrapper)
    Python bindings for the AprilTag fiducial detector via OpenCV's cv2cg, used in a Robot Perception course. A fork, not original work.
  2. slam localization
    ORB-SLAM2 (build + experiment)
    Built and ran the canonical ORB-SLAM2 stack (Mur-Artal et al.) on a custom monocular dataset during a Robot Perception course. Forked the upstream.
  3. deep learning
    Caffe PoseNet (fork)
    Forked the Caffe implementation of PoseNet (regress 6-DOF camera pose directly from a single RGB image). Pre-PyTorch era of CV research.
  4. deep learning
    TensorFlow PoseNet (port)
    Port of PoseNet from Caffe to TensorFlow, also during the Robot Perception course. Same model, friendlier toolchain.
  5. deep learning
    PyTorch Tutorial (fork of 莫烦Python)
    Forked MorvanZhou's PyTorch tutorial repo while learning PyTorch in 2019. The fork mostly served as my notebook running history.
  6. systems cpp
    SD Embedded — Mini-car firmware
    Embedded C/C++ firmware for a mini self-driving car. PWM, sensor sampling, basic PID — the bottom of the SDC stack.
  7. data engineering
    Codes Everyday
    A self-imposed daily-coding challenge in 2019 — write or read code every day, log the result. Mostly LeetCode-flavored Python.
  8. data engineering
    Chrome Power Toys (community guide)
    A community-driven Markdown guide to Chrome power-user features. Not coursework — a 2025 weekend writeup that never grew legs.