Optical Character Recognition Project

Conducted at AI Lab - MTI Technology Vietnam

  • Time: June 2020 - September 2020.
  • Role: AI Engineer Intern
  • Description: Generating data (drug prescription templates), labeling the data, doing Data Augmentation for OCR Project.
  • Result: Finished all the tasks above.

Image Super - Resolution using Deep Learning Techniques

Graduation Thesis (Defense Grade: 9.9/10.0)

https://github.com/luke-dinh/Image-Super-Resolution

  • Time: Jan 2021 - June 2021.
  • Role: Individual, instructed by Dr. Kiet T. Vo.
  • Description: A Deep Learning approach in Single Image Super Resolution field that can create a new image with higher resolution from a low resolution counterpart. I have read about Generative Adversarial Network (GAN) and constructed a model to solve the problem. The new model is called modified Super-Resolution Generative Adversarial Network (mSRGAN).
  • Result: The new model surpasses the same GAN-based and other models in the same problem with higher PSNR and SSIM results and lower the number of total parameters.
  • Language: Python + PyTorch/Tensorflow.

Object Recognition using Convolutional Neural Network

https://github.com/luke-dinh/Convolutional-Neural-Network

  • Time: April 2020 – May 2020.
  • Role: Leader, Programming.
  • Description: Our team’s small project in the course “Artificial Intelligence in Control Engineering”. I constructed the architecture of 3 types of CNN: VGG16, VGG19, ResNet50 and used them for Object Recognition Problem.
  • Target: Comparing the accuracy of results trained by three models above.
  • Language: Python + OpenCV, Keras/Tensorflow.

Bottle - Checking

Our team’s final project in Computer Vision classes

https://github.com/luke-dinh/Bottle-checking

  • Time: Sep 2019 – Dec 2019
  • Role: Member of a team of 4 members, in which, I were responsible for building algorithms and theory.
  • Description: The aim of the project was to construct a system that could check whether the labels and the liquid level of the product are qualified or not. The project included Image Processing and some Machine Learning techniques such as Otsu Thresholding, Object detection, etc.
  • Result: The system was able to check the labels and the liquid level corectly.
  • Language: Python.

Digital Image Processing with MATLAB

A small project when I worked at my teacher’s lab at HCMC University of Technology

https://github.com/luke-dinh/Image-Processing-with-Matlab

  • Time: Sep 2019 - Oct 2019.
  • Role: Individual.
  • Description: I applied some Image Processing techniques in MATLAB to process image and create a new filter based on knowledge learned from the course “Digital Image Processing”.
  • Target: New images after using traditional filters and constructing a new filter.
  • Language: MATLAB.

LZW Image Compression Technique

https://github.com/luke-dinh/LZW-Technique

  • Time: Oct 2019 - Nov 2019.
  • Role: Leader, Algorithm Developer, Programming.
  • Description: This was my final project with 2 classmates in the course “Digital Image Processing”. Our work was developing and programming the LZW Image (Data) Compression Technique using Python.
  • Target: A program that could compress the image for easier transmission.
  • Language: Python.

Go to the Home Page