data:image/png;base64,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data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAeAAAAFoCAYAAACPNyggAAAOhUlEQVR4Xu3VwQkAAAjEMN1/abewn7jAQRC64wgQIECAAIF3gX1fNEiAAAECBAiMAHsCAgQIECAQCAhwgG6SAAECBAgIsB8gQIAAAQKBgAAH6CYJECBAgIAA+wECBAgQIBAICHCAbpIAAQIECAiwHyBAgAABAoGAAAfoJgkQIECAgAD7AQIECBAgEAgIcIBukgABAgQICLAfIECAAAECgYAAB+gmCRAgQICAAPsBAgQIECAQCAhwgG6SAAECBAgIsB8gQIAAAQKBgAAH6CYJECBAgIAA+wECBAgQIBAICHCAbpIAAQIECAiwHyBAgAABAoGAAAfoJgkQIECAgAD7AQIECBAgEAgIcIBukgABAgQICLAfIECAAAECgYAAB+gmCRAgQICAAPsBAgQIECAQCAhwgG6SAAECBAgIsB8gQIAAAQKBgAAH6CYJECBAgIAA+wECBAgQIBAICHCAbpIAAQIECAiwHyBAgAABAoGAAAfoJgkQIECAgAD7AQIECBAgEAgIcIBukgABAgQICLAfIECAAAECgYAAB+gmCRAgQICAAPsBAgQIECAQCAhwgG6SAAECBAgIsB8gQIAAAQKBgAAH6CYJECBAgIAA+wECBAgQIBAICHCAbpIAAQIECAiwHyBAgAABAoGAAAfoJgkQIECAgAD7AQIECBAgEAgIcIBukgABAgQICLAfIECAAAECgYAAB+gmCRAgQICAAPsBAgQIECAQCAhwgG6SAAECBAgIsB8gQIAAAQKBgAAH6CYJECBAgIAA+wECBAgQIBAICHCAbpIAAQIECAiwHyBAgAABAoGAAAfoJgkQIECAgAD7AQIECBAgEAgIcIBukgABAgQICLAfIECAAAECgYAAB+gmCRAgQICAAPsBAgQIECAQCAhwgG6SAAECBAgIsB8gQIAAAQKBgAAH6CYJECBAgIAA+wECBAgQIBAICHCAbpIAAQIECAiwHyBAgAABAoGAAAfoJgkQIECAgAD7AQIECBAgEAgIcIBukgABAgQICLAfIECAAAECgYAAB+gmCRAgQICAAPsBAgQIECAQCAhwgG6SAAECBAgIsB8gQIAAAQKBgAAH6CYJECBAgIAA+wECBAgQIBAICHCAbpIAAQIECAiwHyBAgAABAoGAAAfoJgkQIECAgAD7AQIECBAgEAgIcIBukgABAgQICLAfIECAAAECgYAAB+gmCRAgQICAAPsBAgQIECAQCAhwgG6SAAECBAgIsB8gQIAAAQKBgAAH6CYJECBAgIAA+wECBAgQIBAICHCAbpIAAQIECAiwHyBAgAABAoGAAAfoJgkQIECAgAD7AQIECBAgEAgIcIBukgABAgQICLAfIECAAAECgYAAB+gmCRAgQICAAPsBAgQIECAQCAhwgG6SAAECBAgIsB8gQIAAAQKBgAAH6CYJECBAgIAA+wECBAgQIBAICHCAbpIAAQIECAiwHyBAgAABAoGAAAfoJgkQIECAgAD7AQIECBAgEAgIcIBukgABAgQICLAfIECAAAECgYAAB+gmCRAgQICAAPsBAgQIECAQCAhwgG6SAAECBAgIsB8gQIAAAQKBgAAH6CYJECBAgIAA+wECBAgQIBAICHCAbpIAAQIECAiwHyBAgAABAoGAAAfoJgkQIECAgAD7AQIECBAgEAgIcIBukgABAgQICLAfIECAAAECgYAAB+gmCRAgQICAAPsBAgQIECAQCAhwgG6SAAECBAgIsB8gQIAAAQKBgAAH6CYJECBAgIAA+wECBAgQIBAICHCAbpIAAQIECAiwHyBAgAABAoGAAAfoJgkQIECAgAD7AQIECBAgEAgIcIBukgABAgQICLAfIECAAAECgYAAB+gmCRAgQICAAPsBAgQIECAQCAhwgG6SAAECBAgIsB8gQIAAAQKBgAAH6CYJECBAgIAA+wECBAgQIBAICHCAbpIAAQIECAiwHyBAgAABAoGAAAfoJgkQIECAgAD7AQIECBAgEAgIcIBukgABAgQICLAfIECAAAECgYAAB+gmCRAgQICAAPsBAgQIECAQCAhwgG6SAAECBAgIsB8gQIAAAQKBgAAH6CYJECBAgIAA+wECBAgQIBAICHCAbpIAAQIECAiwHyBAgAABAoGAAAfoJgkQIECAgAD7AQIECBAgEAgIcIBukgABAgQICLAfIECAAAECgYAAB+gmCRAgQICAAPsBAgQIECAQCAhwgG6SAAECBAgIsB8gQIAAAQKBgAAH6CYJECBAgIAA+wECBAgQIBAICHCAbpIAAQIECAiwHyBAgAABAoGAAAfoJgkQIECAgAD7AQIECBAgEAgIcIBukgABAgQICLAfIECAAAECgYAAB+gmCRAgQICAAPsBAgQIECAQCAhwgG6SAAECBAgIsB8gQIAAAQKBgAAH6CYJECBAgIAA+wECBAgQIBAICHCAbpIAAQIECAiwHyBAgAABAoGAAAfoJgkQIECAgAD7AQIECBAgEAgIcIBukgABAgQICLAfIECAAAECgYAAB+gmCRAgQICAAPsBAgQIECAQCAhwgG6SAAECBAgIsB8gQIAAAQKBgAAH6CYJECBAgIAA+wECBAgQIBAICHCAbpIAAQIECAiwHyBAgAABAoGAAAfoJgkQIECAgAD7AQIECBAgEAgIcIBukgABAgQICLAfIECAAAECgYAAB+gmCRAgQICAAPsBAgQIECAQCAhwgG6SAAECBAgIsB8gQIAAAQKBgAAH6CYJECBAgIAA+wECBAgQIBAICHCAbpIAAQIECAiwHyBAgAABAoGAAAfoJgkQIECAgAD7AQIECBAgEAgIcIBukgABAgQICLAfIECAAAECgYAAB+gmCRAgQICAAPsBAgQIECAQCAhwgG6SAAECBAgIsB8gQIAAAQKBgAAH6CYJECBAgIAA+wECBAgQIBAICHCAbpIAAQIECAiwHyBAgAABAoGAAAfoJgkQIECAgAD7AQIECBAgEAgIcIBukgABAgQICLAfIECAAAECgYAAB+gmCRAgQICAAPsBAgQIECAQCAhwgG6SAAECBAgIsB8gQIAAAQKBgAAH6CYJECBAgIAA+wECBAgQIBAICHCAbpIAAQIECAiwHyBAgAABAoGAAAfoJgkQIECAgAD7AQIECBAgEAgIcIBukgABAgQICLAfIECAAAECgYAAB+gmCRAgQICAAPsBAgQIECAQCAhwgG6SAAECBAgIsB8gQIAAAQKBgAAH6CYJECBAgIAA+wECBAgQIBAICHCAbpIAAQIECAiwHyBAgAABAoGAAAfoJgkQIECAgAD7AQIECBAgEAgIcIBukgABAgQICLAfIECAAAECgYAAB+gmCRAgQICAAPsBAgQIECAQCAhwgG6SAAECBAgIsB8gQIAAAQKBgAAH6CYJECBAgIAA+wECBAgQIBAICHCAbpIAAQIECAiwHyBAgAABAoGAAAfoJgkQIECAgAD7AQIECBAgEAgIcIBukgABAgQICLAfIECAAAECgYAAB+gmCRAgQICAAPsBAgQIECAQCAhwgG6SAAECBAgIsB8gQIAAAQKBgAAH6CYJECBAgIAA+wECBAgQIBAICHCAbpIAAQIECAiwHyBAgAABAoGAAAfoJgkQIECAgAD7AQIECBAgEAgIcIBukgABAgQICLAfIECAAAECgYAAB+gmCRAgQICAAPsBAgQIECAQCAhwgG6SAAECBAgIsB8gQIAAAQKBgAAH6CYJECBAgIAA+wECBAgQIBAICHCAbpIAAQIECAiwHyBAgAABAoGAAAfoJgkQIECAgAD7AQIECBAgEAgIcIBukgABAgQICLAfIECAAAECgYAAB+gmCRAgQICAAPsBAgQIECAQCAhwgG6SAAECBAgIsB8gQIAAAQKBgAAH6CYJECBAgIAA+wECBAgQIBAICHCAbpIAAQIECAiwHyBAgAABAoGAAAfoJgkQIECAgAD7AQIECBAgEAgIcIBukgABAgQICLAfIECAAAECgYAAB+gmCRAgQICAAPsBAgQIECAQCAhwgG6SAAECBAgIsB8gQIAAAQKBgAAH6CYJECBAgIAA+wECBAgQIBAICHCAbpIAAQIECAiwHyBAgAABAoGAAAfoJgkQIECAgAD7AQIECBAgEAgIcIBukgABAgQICLAfIECAAAECgYAAB+gmCRAgQICAAPsBAgQIECAQCAhwgG6SAAECBAgIsB8gQIAAAQKBgAAH6CYJECBAgIAA+wECBAgQIBAICHCAbpIAAQIECAiwHyBAgAABAoGAAAfoJgkQIECAgAD7AQIECBAgEAgIcIBukgABAgQICLAfIECAAAECgYAAB+gmCRAgQICAAPsBAgQIECAQCAhwgG6SAAECBAgIsB8gQIAAAQKBgAAH6CYJECBAgIAA+wECBAgQIBAICHCAbpIAAQIECAiwHyBAgAABAoGAAAfoJgkQIECAgAD7AQIECBAgEAgIcIBukgABAgQICLAfIECAAAECgYAAB+gmCRAgQICAAPsBAgQIECAQCAhwgG6SAAECBAgIsB8gQIAAAQKBgAAH6CYJECBAgIAA+wECBAgQIBAICHCAbpIAAQIECAiwHyBAgAABAoGAAAfoJgkQIECAgAD7AQIECBAgEAgIcIBukgABAgQICLAfIECAAAECgYAAB+gmCRAgQICAAPsBAgQIECAQCAhwgG6SAAECBAgIsB8gQIAAAQKBgAAH6CYJECBAgIAA+wECBAgQIBAICHCAbpIAAQIECAiwHyBAgAABAoGAAAfoJgkQIECAgAD7AQIECBAgEAgIcIBukgABAgQICLAfIECAAAECgYAAB+gmCRAgQICAAPsBAgQIECAQCAhwgG6SAAECBAgIsB8gQIAAAQKBgAAH6CYJECBAgIAA+wECBAgQIBAICHCAbpIAAQIECAiwHyBAgAABAoGAAAfoJgkQIECAgAD7AQIECBAgEAgIcIBukgABAgQIHLFxAWmhEwHPAAAAAElFTkSuQmCChttp://beej.us/blog/data/convolution-image-processing/- select a small colored image The example is 50 x 50 =2500 pixels (the larger the image the longer the processing time is).
- set this image as a costume.
- in SETTINGS set Turbo mode ON. Shift key + Settings to set ON "JIT compiler support"
- download Snap! Pixel library.
- make a Variable for a list named PIX50 (any name of your choice)
- run PIXEL OF CURRENT COSTUME to extract the list of RGB values of the 2500 pixels of selected costume.
- extract the first item of the list (=R) using FMAP. It gives a list of 2500 x 1 values
- make a copy of the first costume. Keep the name given (no need to change it)
- run CONVOL 50 6 (C for convolution)
- it retuns a "new " costume modified.
- what is the parameter 6 ?
- open the pink user block FILTERS. Each line contains the definition of a FEATURE map used to modify the Initial image by "convolution".
- for each fature is gien a name, a number and a list of 9 numbers.: sharpen, darken, blur . "6" is for "Line detection".
- result of each convolutional feature is given as a new costume.
- how is got this result ?
- for each item of PIX50 is created a list of the 9 pixel value around the selected pixel.
- this list is multiplied by the FEATURE list. It gives a new RGB pixel value. New list once created is "loaded into blm50(2) costume using the PIxel library block UPDATE CURRENT COSTUME WITH LIST
- test each feature provided and create your own effect.
- the original size of pixel table is kept. But for real application the role of this convolutional layer is to reduce the size of tables used by the Neural network . To do so only half number of items is selected for feature filtering. Thus , the size of Input table is reduced to one half ie 625 items compared to 2500 !
- to visualize a convolution list of pixels feed it into a Costume of same size 50x50 like blm50reports a list of all pixels in the given costume. Each pixel is represented by a 4-item sublist of RGBA valuespixelspixels
width
heightheightcostume var i,
pixels = [],
src = costume.contents.getContext('2d').getImageData(
0,
0,
costume.width(),
costume.height()
);
for (i = 0; i < src.data.length; i += 4) {
pixels.push(new List([
src.data[i],
src.data[i + 1],
src.data[i + 2],
src.data[i + 3]
]));
}
return new List(pixels);overwrite the pixels in the given costume with another list of pixels.
Caution! This operation directly changes the costume. Make sure to use a copy in case you would like to revert changes later.reports the sprite's actual current costume, which may or may not be part of its wardrobe e.g. if it is a copyreports a copy of the given costume. This is especially useful when modifying a costume, so you can keep the original arounddisplay the given bitmap in a copy of the current costume (so you can switch back to the original costume again)Blur9111111111Edge detect939-3-321-3090Sharpen9111111111Darken6000060000Lighten90000120000Line detection1-1-1-1-18-1-1-1-1V lines12-1-1-12-1-1-12Sharpen Int 1-1-1-1-19-1-1-1-1Gauss blur8010141010Edge Lapalcian 10-10-14-10-10