Wednesday, 6 November 2013

I have pushed my college project on classifying facial expressions using neural networks to github.

https://github.com/lawlessc/face-classifier-trainer

Tuesday, 6 November 2012

Git



My first try of committing a project to bitbucket using git, went smooth enough. I think i will use this more for personal projects and may put my thesis project up on a github public repo after i clean up the code a bit.

Tuesday, 17 July 2012

Working with Libgdx

Have decided to start working with Libgdx to see if i can build a half decent game for the Android platform. The great thing about it is i can run programs on the desktop in Java first without having to use an emulator. Being in Java it's all very easy to code, with the exception that i now have to learn about OpenGL again and how to use shaders, which has been slow to progress.

If i figure out how to use shaders (maybe only partially how they work) it opens up a lot more options  eye candy for making game look better  such as water effects, bloom shaders and cell shaders.

The Libgdx documentation itself seems to be a bit sparse in places as the developers seem to be far more focused on fixing bugs and rapidly developing new features.

Thursday, 12 April 2012

Metasploit

Just tried out Metasploit for the first time, using its 7 day trial. Have to say it really does wrap everything into neat and pretty reports.

Sunday, 25 March 2012

Expression detection using neural networks in OpenCv



The expression classifiers work well for discerning between when a person is smiling and not smiling.For other emotions they perform relatively poorly. I believe this is due to a lack of data, its much harder to find images of people with facial expressions other than smiling.

For discerning between smiling with teeth baring and an angry expression with  teeth baring it performs surprisingly well, compared with how poorly it performs for discerning between fear , surprise and disgust. Perhaps its because these overlap?



All of the networks are using the pixel values of the images as inputs, face images are being scaled to 50x50 images which gives all the classifiers input sizes of 2500, hidden layers are between 9 and 12 neurons in size, and the output layer consists of 2 neurons one for a true value and one for a false value.

Monday, 19 March 2012

Sunday, 11 March 2012

Expression Detection

Seems to partially work, i believe i need more data for training the classifier and more inputs.

This was wrong, it was a completely fluke this worked at all as i had used far to many nodes in the hidden layer