Predspracovanie obrazu pre optické rozpoznávanie ... - TUKE
Predspracovanie obrazu pre optické rozpoznávanie ... - TUKE
Predspracovanie obrazu pre optické rozpoznávanie ... - TUKE
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
FEI TU v Košiciach Diplomová práca List č. 85<br />
in<br />
in<br />
in<br />
1,1<br />
1,2<br />
1,3<br />
in<br />
in<br />
in<br />
2,1<br />
2,2<br />
23<br />
in<br />
in<br />
in<br />
3,1<br />
3,2<br />
3,3<br />
k<br />
2,1<br />
k<br />
Obr. 46: Schematical drawing of a 2D image filter.<br />
in1,1<br />
in1,2<br />
in1,3<br />
in2,1<br />
in2,2<br />
in2,3<br />
in3,1<br />
in3,2<br />
in3,3<br />
1,3<br />
k<br />
3,1<br />
h 1<br />
h 2<br />
h 3<br />
out<br />
out1<br />
out2<br />
Obr. 47: Schematical drawing of neural network<br />
- the output can be one (or more) pixels<br />
- the signal flows from the input to the output<br />
- the input and output are connected with lines which have weights<br />
So a very similiar forward-feed neural network to the 2D filter can look like<br />
one shown on the picture 47. I’ve chosen the parameters of used neural ne-<br />
twork:<br />
- type of neural network: forward-feed full-connection neural network<br />
- learning method: standart error back-propgation<br />
- input layer of N × N neurons<br />
- hidden layer of M neurons<br />
- output layer of 2 neurons (one for direct level of shade and one for<br />
inverted level of shade)