Name : Ngap Wei Tham
Date of Birth :-
Marital Status : single
Open source projects developed by Qt5(similar vision use opencv too):
1 : similar_vision : This project use image hash to find similar images and provide you decent user interfaces.
To avoid gui freezing and speed things up, I leverage the power of multi-thread.
2 : QImageScraper : A desktop app which could scrape images from Google search, Bing search, Yahoo search,
Adobe stock, deposit photo, dream time, shutter stock.
3 : QRenamer : A gui tool to help you rename a lot of files, developed by Qt5(c++).
4 : LightDetector : A cross-platform object recognition app(by MobileNetSSD) which could run on windows, mac,
linux and android.
Commerical projects
1 : EyediusPremium : A security app which could come with various functions.
a. Object detection with various alarm settings
b. Age-gender detection
c. Face recognition with various alarm settings
d. Multiple object tracking
e. Various way to communications with the app by network
f. etc
2. Patent in image classification : Classify this page contains text or image by cnn, trained
by Pytorch.
Another open source computer vision projects :
1 : tiny_cnn_test : This project use cnn to classify car vs non car.
2 : CBIR : Use opencv and armadillo to implement bag of visual words(BOVW).I use BOVW to find similar images,
but not classification or recognition in this project(test on ukbench).
3 : tic_tac_toe : Use opencv(c++) to recognize X and O, very simple but fun.
4 : python_deep_learning : Deep learning exercise and project developed by python, it
show you how to
A : Process imagenet data by python with SIMD and parallel
B : Neural style transfer
C : Super resolution
D : Semantic segmentation
5 : ANPR2 : Use opencv(c++) to implement automatic number plate recognition, this project build a complete
pipeline of ANPR, license plate detect->perspective transform->characters extraction->OCR. Accuracy is far form
perfect, this could be improved with more data and more advance technique like deep learning(I did not know deep
learning when I develop this project).
6 : Semantic
segmentation by LinkNet : This is the first working example of LinkNet implemented by
PyTorch(unless I cannot find it by DuckDuckGo), for more details, please check this post.
7 : stitching : small, simple demo of image stitching by OpenCV3.
Deep image homography : Surpass the accuracy of the paper 3.5 times with smaller,
faster cnn model. Developed by torch and able to import by opencv dnn module.
8:
9 : Person detect with mxnet : Show you how to perform object detection with mxnet,
opencv and c++.
10 : Train a detector based on yolo v3 by custom data : Show you how to train a face,
person detector by gluoncv
Kaggle competition :
Top 6% in cat vs dog
Top 7% in the nature conservancy Fisheries Monitoring
You can check the results at here.
Third party c++ project I contributed:
1 : mlpack
What I contribute to this communities
a : Develop layers of sparse autoencoder by neural network module of mlpack
b : Bug fix and refactor
c : Provide train test split function
d : Provide function to view the results of autoencoder
e : Mentor of GSOC(google summer of codes) in 2016
f : Provide command tool for softmax classifier
g : Develop a csv parser for mlpack
2 : opencv_contrib
a : Contribute a new module--img_hash(image hash) to opencv, it is one of the
contribute model of opencv
Education :
National Cheng Kung University
Master of Electrical Engineering, 2009/09~2011/09
(Master Thesis: Low Complexity Cross-Based Local Stereo- Matching for Fast Depth
Estimation)
National Chiao Tung University
Bachelor of Electrical Engineering, 2004/09~2008/09
PyimageSearch Gurus
Finished the courses of PyimageSearch Gurus in 03/22/2016, start from 07/06/2015
Skills
1. Programming
GitHub Account : stereomatchingkiss
Main Languages : c++
Secondary langauges : Javascript, Python,qml
Libraries : Boost, openCV, Qt5, dlib, Keras, mlpack, intelTBB, standard
library of c++11/14, PyTorch, torch
Concepts of multithread
CMake, qmake(they are make file generator)
2. Digital Image Processing/Computer vision/Deep learning
3. Languages
Mandarin(native language)
English
Cantonese
4. Contact
Email :-
mobile : -
Skype name : thamngapwei
