UCF REU: Weeks 1 & 2. Gradient Code Gradient Direction of the Gradient: Calculating theta.

UCF REU:Weeks 1 & 2

Gradient

Code

Gradient

Direction of the Gradient:Calculating theta

Picking a threshold for the gradient map

Subtract the gradient from the original image display

it

Detailing oncamera

Edge Detection

Sobel approximation to the derivative

Filters

Lapacian of Gaussian (LoG)

Canny Edge DetectorGaussian filter to smooth image

1st derivative kernel to detect edges

Non-maximal suppression

Median filter

• B = medfilt2(im) does a median filtering of the image im in two dimensions.

• Each output pixel contains the median value in the 3 x 3 neighborhood around the corresponding pixel in the input image. It pads the image with zeros on the edges, so the median values for points on the edge of the image may appear darker.

• Median filtering is most often used to reduce "salt and pepper" noise.

• A median filter is more effective than convolution when the goal is to simultaneously reduce noise and preserve edges.

Harris Corner Detector

Shifting the window in any direction should yield a large change in appearance

-aka what happens at a corner!

Code for Harris Corner

Optical Flow (Lucas Kanade with

pyramids)

Shows the vectors of motion.

Optical Flow (Liu)

• Found YouTube clip of Olympic diver: https://www.youtube.com/watch?v=djou2oLUeuo

• Clipped the video into a 3 second clip of just the action of diving

• Converted the clipped video into a sequence of jpg images• Example:

• Made a loop in demo flow to run through the sequence of diving images

• Took the optical flow images and converted them into an animated gif:Olympic Diver: http://makeagif.com/I27cPy

Scale Invariant Feature Transform (SIFT)

Code (for one image)

Code to collect all images

Have to do these 2 pieces of code for every 15 categories

Input images

Basic matching using SIFT points

Result

Code

Adaboost

SVM (Support Vector Machine)

Pick classifier where the distance between the support vectors and the linear classifier is maximized.

Bag Of Words

Project Preference 1Project: Multimodal data analysis for the

detection of Attention Deficit Hyperactive Disorder

Mentor: Soumyabrata Dey

Began his reading most recent paper on the project

Have taken discrete systems which has graph theory which will help because: looking at the brain as a series of nodes and edges

Project Preference 2Project: Deep Tracker

Mentor: Afshin Dehghan

I know python

Looked at papers on deep learning

Project Preference 3Project: Action Recognition in Temporally

Untrimmed videos

Mentor: Amir R. Zamir

Have a solid grasp of frame work

Very interested in topic