Necessary source files and resources for the Iowa State Spring 2016 OpenCV Lab Jam
Make sure Python 2.7 is installed on your machine.
Make sure you copy cv2.pyd from the x86 directory under OpenCV Make sure you have numpy version 9 at minimum
Follow this link to install the OpenCV Libraries in Windows.
Follow this link to install the OpenCV libraries in Linux.
Follow this link to install the OpenCV libraries in OS X.
If you are using the yum package manager (Fedora or other relative distros), the following command should successfully install OpenCV:
sudo yum install numpy opencv*
This may not give you access to the latest OpenCV version like building from source, so there may be compatibility issues involved. It's worth a shot to try.
You can use whatever editor you want for this. I would recommend PyCharm. Visual Studio and XCode are also great options.
Wait until the lab jam to continue
Try running color-filtering/color.py and see what happens. Follow this link to learn more about HSV image parameters.
Make sure you hit escape to exit the program. Don't hit X in the corner of the window. Some GUI features are messed up (at least on linux)
What we are doing is a very simple image processing technique known as thresholding. Some parts of the image are within the desired parameters, and are thresholded to a 1 (White pixel) while those outside our desired parameters are thrown out and set to 0 (black pixels.) This is very helpful for finding parts of an image that are a certain color. You are seeing a bitmasked output by default, so the original image is overlaid with the "white" or high thresholded pixels.
First Step: Have your program hardcoded to select only one cartridge on startup.
Free cookie: Using OpenCV contours, draw a rectangle around whichever pokemon cartridge your settings are selecting.
Now change to the face-detect directory and try out face detection using:
python face.py <image name> haarcascade_frontalface_default.xml
Here's some good info on haar cascade classifiers. Face detection is a very advanced subject and using machine learning to recognize faces has been proven to be an effective solution. We are utilizing training done by professionals to help us.
Thankless Task: Use haarcascade_eye.xml as a second classifier to detect eyes in the images given. Draw rectangles around the both the eyes and faces (this may not work perfectly, just try your best to find settings that get the most eyes.)
Fun Extra: All of the premade haarcascade classifier files are found where you installed opencv in opencv/data/haarcascades/. If you so choose, you can pick one of them and your own image and try to find anything that has been trained.
Now that you have a few handy tools under your belt, feel free to go wild with this last step!
Run video-stream/video.py to see the output from your webcam. You can access the frames in real time. If you don't have a webcam on your device, just download any video format file and specify the path as the argument of cv2.VideoCapture().
Do something awesome: With your basic toolset, do some frame by frame processing . A good starting point is to run the face detection on your video feed and see if it can track your face!
More OpenCV Tools: Background Subtraction, Circle Detection, Optical Flow, Edge Detection