My Projects

This page introduces some of my projects.

Description: Daisykit is an open source AI toolkit, aiming to provide an easy way to use AI in software projects on multiple platform. Now we have C++, Python and Android demos.
Techstack (Current): OpenCV, NCNN, C++, Bindings for Java and Python
My role: Founder and Developer
Project Website: https://daisykit.org/
Github: Repos from https://github.com/Daisykit-AI/daisykit
Demo: Repos from https://youtu.be/zKP8sgGoFMc

Description: VIA Project is an open source for self-driving cars, including source code and tutorials for learning image processing and computer vision by building autonomous vehicles, VIA Simulation - a highly customizable simulation for self-driving car testing and VIA Makerbot - PCB for your mini self-driving car.
My role: Lead of AI content.
Project Website (In Vietnamese): https://via.makerviet.org/
Github: Repos from https://github.com/makerhanoi
Demo: https://youtu.be/N0HzDScBAxo

Description: VN AIDr project is an open source medical image processing solution for AI learners and hobbyists. The core of this project is a medical image processing platform using client-server architecture. On this platform, we design, train and evaluate machine learning models and algorithms for medical image analysis and medical report autocompletion.
My role: Leader + Build deep learning models.
Project Website: https://vnopenai.github.io/ai-doctor/
Github: https://github.com/VNOpenAI/vn-aidr

Description: This project employs BlazePose - a SOTA model for real-time human pose detection, a single frame action recognition model, and some signal processing algorithms to count pushups from a video stream or webcam.
My role: Leader + Build BlazePose model.
Github: https://github.com/VNOpenAI/pushup-counter-app
Demo: https://www.youtube.com/watch?v=pm2mGmgwSZo

Description: An ADAS system that uses Jetson Nano as the hardware with four main functions: forward collision warning, lane departure warning, traffic sign recognition and overspeed warning. I trained and optimized three deep neural networks to run simultaneously on Jetson Nano (CenterNet-ResNet18 for object detection, U-Net for lane line segmentation and ResNet-18 for traffic sign classification). I won The NVIDIA Jetson Project Of the Month Prize with this project and received a Jetson AGX Xavier Developer Kit.
Languages: Python, C++.
Technologies: Tensorflow, Pytorch, TensorRT, object detection (CenterNet), segmentation (U-Net) and classification (ResNet-18)
Github: https://github.com/vietanhdev/car-smart-cam
Blog posts:

Description: This project helped my group achieving The First Prize in FPT Self-driving Car challenge 2020 - University round at Hanoi University of Science and Technology. In this challenge, we built a ROS node with deep neural networks to control a autonomous car in a simulation environment. My group used E-Net and U-Net for lane segmentation and customized FaceBoxes for a light-weight traffic sign detection networks. I was the leader of this group, worked on the architecture, semantic segmentation model and car controlling.
Languages: Python.
Technologies: ROS (Robot Operating System), OpenCV, segmentation models (U-Net, ENet), object detection (FaceBoxes), Docker.
Github: https://github.com/vietanhdev/autonomous-car-2020
Blog post: https://aicurious.io/posts/chung-toi-da-xay-dung-xe-tu-hanh-tren-gia-lap-the-nao/

Description: I led a 4-member team to build a paper streaming solution for remote education. Our project got Second Place in SoICT - IBM Hackathon 2020 at HUST. I was responsible for algorithms to capture drawing strokes and handwritten text from a phone camera, filter and create video streaming.
Languages: Python.
Technologies: ARUCO, Color-based segmentation, ReactJS, NodeJS, CI-CD.
Github: https://github.com/vietanhdev/paper_stream
Blog post: https://aicurious.io/posts/xay-dung-giai-phap-stream-giay-viet-IBM-Hackathon-2020/
Demo: https://www.youtube.com/watch?v=pmRSfHSfrco

Description: A modified Deep Head Pose model by adding new backbones (ShuffleNetV2, EfficientNet), and face landmark estimation.
Languages: Python.
Technologies: Deep Head Pose, ShuffleNetV2, EfficientNet, Landmark Estimation.
Demo: https://youtu.be/SNHnsuNkBkQ
Github: https://github.com/vietanhdev/deep-head-pose-2

Description: A camera desktop application with funny face decorations and filters. In this project, I also integrated LBP Cascade and Haar Cascade models trained by myself to detect faces. This project supports both Linux and Windows.
Languages: C++.
Technologies: Qt, OpenCV, face detection and alignment algorithms.
Github: https://github.com/vietanhdev/facecam

Description: This is a group project for FPT self-driving car challenge 2018. We used Watershed and Floodfill algorithms for lane segmentation, a color-based method for traffic sign detection and HOG + SVM for sign classification. I worked on the architecture, algorithms for lane segmentation, a method for traffic sign detection and car control algorithm.
Languages: C++.
Technologies: OpenCV, Image processing, Color-based Object detection, Traffic Sign detection using HOG SVM.
Github: https://github.com/vietanhdev/autonomous-car-2018
Demo: https://www.youtube.com/watch?v=Obv53r7UV34.

Description: We built a completed store management solution with a lot of features such as billing, management UI for bills, warehouse, suppliers, customers, finance and employees. I worked on the desktop application.
Languages: Desktop App: Javascript, HTML CSS; Server: Java
Technologies: Desktop App: Electron; Server: Java Spring Boot.
Github: https://github.com/vietanhdev/e-store-manager
Demo: Youtube Video.

Description: My funny IoT project for our environment. This trash bin can classify bottles and other kinds of garbage into different compartments. I finished this project in only 1 day. Most of the time was spent for the mechanical part, sensors, servo. I integrated and run a pre-trained SSD-MobileNet V2 model on a Raspberry Pi computer to detect bottles.
Technologies: Desktop App: Electron; Server: Java Spring Boot.
Demo: https://www.youtube.com/watch?v=s1KHGU8nuBM
Blog post (Vietnamese): Thiết kế thùng rác thông minh tự phân loại rác với Raspberry Pi 3.

Description: Online BattleShip game using ReactJS for frontend, Flask for backend and Websocket for realtime communication. We also implemented a chatting system and friend list. I worked mostly on frontend part (ReactJS).
Github: https://github.com/vietanhdev/online-battleship