Skip to content
This repository has been archived by the owner on Sep 26, 2023. It is now read-only.

Commit

Permalink
some reworkings of window code
Browse files Browse the repository at this point in the history
  • Loading branch information
@harthur
harthur committed Jun 6, 2012
1 parent 20ce119 commit caad1da
Showing 1 changed file with 116 additions and 102 deletions.
218 changes: 116 additions & 102 deletions kittydar.js
View file
Expand Up @@ -8,130 +8,152 @@ var net = new brain.NeuralNetwork().fromJSON(network);

var threshold = 0.9;

exports.detectCats = function detectCats(canvas) {
var imagedata = canvas;
if (imagedata.data && imagedata.width && imagedata.height) {
// it's actually an imagedata object
canvas = new Canvas(imagedata.width, imagedata.height);
var ctx = canvas.getContext("2d");
ctx.putImageData(imagedata, 0, 0);
}

var width = canvas.width;
var height = canvas.height;

var max = Math.max(width, height)
var scale = Math.min(max, 360) / max;

width *= scale;
height *= scale;
var kittydar = {
detectCats: function(canvas) {
var imagedata = canvas;
if (imagedata.data) {
// it's actually an ImageData object, put in canvas
canvas = new Canvas(imagedata.width, imagedata.height);
var ctx = canvas.getContext("2d");
ctx.putImageData(imagedata, 0, 0);
}

canvas = resizeCanvas(canvas, width, height);
var max = Math.max(canvas.width, canvas.height)
var scale = Math.min(max, 360) / max;
var width = Math.floor(canvas.width * scale);
var height = Math.floor(canvas.height * scale);

var min = 48; // starting window size
canvas = resizeCanvas(canvas, width, height);

var cats = [];
for (var size = min; size < max; size += 12) {
cats = cats.concat(detectAtSize(canvas, size, min));
}
var min = 48; // starting window size

cats = cats.map(function(cat) {
return {
x: Math.floor(cat.x / scale),
y: Math.floor(cat.y / scale),
width: Math.floor(cat.width / scale),
height: Math.floor(cat.height / scale),
prob: cat.prob
var cats = [];
for (var size = min; size < max; size += 12) {
cats = cats.concat(this.detectAtSize(canvas, size, min));
}
});

return cats;
}

function detectAtSize(canvas, size, fixed) {
var scale = fixed / size;

var width = Math.floor(canvas.width * scale);
var height = Math.floor(canvas.height * scale);

canvas = resizeCanvas(canvas, width, height);
var ctx = canvas.getContext("2d");
var imagedata = ctx.getImageData(0, 0, width, height);
cats = cats.map(function(cat) {
return {
x: Math.floor(cat.x / scale),
y: Math.floor(cat.y / scale),
width: Math.floor(cat.width / scale),
height: Math.floor(cat.height / scale),
prob: cat.prob
}
});

return cats;
},

detectCatsYourself: function(canvas, callback) {
// for use with Worker threads, callback gets called with imagedata
// object you can send to thread to call detectAtFixed() on
var imagedata = canvas;
if (imagedata.data) {
// it's actually an ImageData object, put in canvas
canvas = new Canvas(imagedata.width, imagedata.height);
var ctx = canvas.getContext("2d");
ctx.putImageData(imagedata, 0, 0);
}

var cats = detectAtFixed(imagedata, fixed, scale);
return cats;
}
var max = Math.max(canvas.width, canvas.height)
var scale = Math.min(max, 360) / max;
var width = Math.floor(canvas.width * scale);
var height = Math.floor(canvas.height * scale);

function isCat(intensities) {
var options = {
"cellSize": 4,
"blockSize": 2,
"blockStride": 1,
"bins": 6,
"norm": "L2"
};

var fts = hog.extractHOGFromIntensities(intensities, options);
var prob = net.run(fts)[0];
return prob;
}
canvas = resizeCanvas(canvas, width, height);

function detectAtFixed(imagedata, fixed, scale) {
console.log("detect at fixed", fixed, scale);
var shift = 6;
var min = 48; // starting window size

var intensities = hog.intensities(imagedata);
var cats = [];
for (var size = min; size < max; size += 12) {
var imagedata = this.resizeToFixed(canvas, size, min);
callback(imagedata);
}
},

var cats = [];
resizeToFixed: function(canvas, size, fixed) {
var scale = fixed / size;

for (var y = 0; y + fixed < imagedata.height; y += shift) {
for (var x = 0; x + fixed < imagedata.width; x += shift) {
var win = getRect(intensities, x, y, fixed, fixed);
var prob = isCat(win);
var width = Math.floor(canvas.width * scale);
var height = Math.floor(canvas.height * scale);

if (prob > threshold) {
cats.push({
x: Math.floor(x / scale),
y: Math.floor(y / scale),
width: Math.floor(fixed / scale),
height: Math.floor(fixed / scale),
prob: prob
});
// resize the image so the fixed size can mimic window size
canvas = resizeCanvas(canvas, width, height);
var ctx = canvas.getContext("2d");
var imagedata = ctx.getImageData(0, 0, width, height);

return imagedata;
},

detectAtSize: function(canvas, size, fixed, callback) {
var imagedata = this.resizeToFixed(canvas, size, fixed);

var cats = this.detectAtFixed(imagedata, fixed, scale);
return cats;
},

isCat: function(intensities) {
var options = {
"cellSize": 4,
"blockSize": 2,
"blockStride": 1,
"bins": 6,
"norm": "L2"
};

var fts = hog.extractHOGFromIntensities(intensities, options);
var prob = net.run(fts)[0];
return prob;
},

detectAtFixed: function(imagedata, fixed, scale) {
// Only detect using a sliding window of a fixed size.
// Take an ImageData instead of canvas so that this can be
// used from a Worker thread.
var intensities = hog.intensities(imagedata);
var shift = 6;
var cats = [];

for (var y = 0; y + fixed < imagedata.height; y += shift) {
for (var x = 0; x + fixed < imagedata.width; x += shift) {
var win = getRect(intensities, x, y, fixed, fixed);
var prob = this.isCat(win);

if (prob > threshold) {
cats.push({
x: Math.floor(x / scale),
y: Math.floor(y / scale),
width: Math.floor(fixed / scale),
height: Math.floor(fixed / scale),
prob: prob
});
}
}
}
return cats;
}
return cats;
}

function getRect(matrix, x, y, width, height) {
function getRect2(matrix, x, y, width, height) {
var square = new Array(height);

for (var i = 0; i < height; i++) {
square[i] = new Array(width);

for (var j = 0; j < width; j++) {
square[i][j] = matrix[y + i][x + j];
}
}
return square;
}

function getIntensities(elements, x, y, size, bins) {
var histogram = zeros(bins);

for (var i = 0; i < size; i++) {
for (var j = 0; j < size; j++) {
var vector = elements[y + i][x + j];
var bin = binFor(vector.orient, bins);
histogram[bin] += vector.mag;
}
function getRect(matrix, x, y, width, height) {
matrix = matrix.slice(y, y + height);
for (var i = 0; i < height; i++) {
matrix[i] = matrix[i].slice(x, x + width)
}
return histogram;
return matrix;
}




function resizeCanvas(canvas, width, height) {
var resizeCanvas = new Canvas(width, height);
var ctx = resizeCanvas.getContext('2d');
Expand All @@ -142,12 +164,4 @@ function resizeCanvas(canvas, width, height) {
return resizeCanvas;
}

function cropAndResize(canvas, x, y, size, resizeTo) {
var resizeCanvas = new Canvas(resizeTo, resizeTo);
var ctx = resizeCanvas.getContext('2d');
ctx.patternQuality = "best";

ctx.drawImage(canvas, x, y, size, size, 0, 0, resizeTo, resizeTo);

return resizeCanvas;
}
module.exports = kittydar;

0 comments on commit caad1da

Please sign in to comment.