# NUUEE 2020 Computer Vision

<h2>
Gesture recognition by using machine Learning
</h2>

The result of gesture recognition is dependent on the effective in skin segmentation process.

Machine learning processing should need huge amounts of training data,
with better and more pure quality of training data you use the better result you get.

* Step :

1. Prepare the training data. In this project, you need picture the target or record a video to get enough amount of training samples.

3. Classify the training data samples in step 1.
Put each sample on the difference folder that names by sample's meaning names.

3. Prepare the Feature extract function code, adjust the extraction parameter and check the feature extract result before training the classify. The more features you get and more difference between each one category of sample, the better quality of classify you get.

* SURF

* BOW (Bag of words)
* Reference [ICCV 2009 Recognizing and Learning Object Categories: Year 2009
](http://people.csail.mit.edu/torralba/shortCourseRLOC/)

* SVM (Support Vector Machine)

<h1>
2020- Fall computer vision
</h1>

* Object Detect
* Face Regonition
* Object Tracking
* CNN ---> Octave CNN
* Oriented Gradients
* Pixels From Light to Dark or Pixels From Dark to Light.
* Semantic Segmentation
* Instance segmentation
* High-level
* Image Classification
* Object detection
* Retrieval
* Robots

* Homework 01
Create a furits image that include banana, sanji, lemon, tomato and purple brussels sprouts.
Design a program that written by opencv-python to classify these five furits depand on its color.
Suggest: using Matpoltlib.
[Source code](https://github.com/shawnlintw/img_processing/tree/develop/python/hw/01_color)

* Cross-Correlation v.s. Convolution
* Laplacian of Gaussian (LoG)
* $G(x,y)=\dfrac{1}{\sqrt{(2\pi)\sigma}} e^{-\dfrac{x^2+y^2}{2\sigma^2} }$
*
* Homework 02
Develop a Canny Edge detection program onto Tkinter GUI. Its functions shows below:
* Load a video file, static image or capture image from camera and display on the source image window then do canny edge detection shows the result on destination window.
* Panorama
* Error= min ($\Sigma_d\Sigma_{x,y}(I(x,y)-I(x+d_x,y+d_y))^2$)

* Homework 03
Panorama :
* Stitch the campus photos (left , center and right) of NUU.

* Homework 04
Moving object tracking :
* Detect and track moving object by using LK optical flow method and Farneback optical flow method, from a video (cam) and observe what's difference between these two methods.