Mercurial > hg > nsaunier > traffic-intelligence

comparison python/cvutils.py @ 680:da1352b89d02 dev

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
classification is working
author Nicolas Saunier <nicolas.saunier@polymtl.ca>
date Fri, 05 Jun 2015 02:25:30 +0200
parents 5505f9dbb28e
children 94b291a5f933
comparison
equal deleted inserted replaced
678:97c305108460 680:da1352b89d02
15 except ImportError: 15 except ImportError:
16 print('Scikit-image library could not be loaded (HoG-based classification methods will not be available)') 16 print('Scikit-image library could not be loaded (HoG-based classification methods will not be available)')
17 skimageAvailable = False 17 skimageAvailable = False
18 18
19 from sys import stdout 19 from sys import stdout
20 from numpy import dot, array, append 20 from numpy import dot, array, append, float32
21 21
22 #import aggdraw # agg on top of PIL (antialiased drawing) 22 #import aggdraw # agg on top of PIL (antialiased drawing)
23 23
24 24
25 cvRed = (0,0,255) 25 cvRed = (0,0,255)
198 '''Returns nFrames images from the video sequence''' 198 '''Returns nFrames images from the video sequence'''
199 from math import floor, log10 199 from math import floor, log10
200 images = [] 200 images = []
201 capture = cv2.VideoCapture(videoFilename) 201 capture = cv2.VideoCapture(videoFilename)
202 if capture.isOpened(): 202 if capture.isOpened():
203 nDigits = int(floor(log10(capture.get(cv2.cv.CV_CAP_PROP_FRAME_COUNT))))+1 203 rawCount = capture.get(cv2.cv.CV_CAP_PROP_FRAME_COUNT)
204 if rawCount < 0:
205 rawCount = firstFrameNum+nFrames+1
206 nDigits = int(floor(log10(rawCount)))+1
204 ret = False 207 ret = False
205 capture.set(cv2.cv.CV_CAP_PROP_POS_FRAMES, firstFrameNum) 208 capture.set(cv2.cv.CV_CAP_PROP_POS_FRAMES, firstFrameNum)
206 imgNum = 0 209 imgNum = 0
207 while imgNum<nFrames: 210 while imgNum<nFrames:
208 ret, img = capture.read() 211 ret, img = capture.read()
230 return fps 233 return fps
231 else: 234 else:
232 print('Video capture for {} failed'.format(videoFilename)) 235 print('Video capture for {} failed'.format(videoFilename))
233 return None 236 return None
234 237
235 def imageBox(img, obj, frameNum, homography, width, height, px = 0.2, py = 0.2, pixelThreshold = 800): 238 def imageBox(img, obj, frameNum, homography, width, height, px = 0.2, py = 0.2, minNPixels = 800):
236 'Computes the bounding box of object at frameNum' 239 'Computes the bounding box of object at frameNum'
237 x = [] 240 x = []
238 y = [] 241 y = []
239 if obj.hasFeatures(): 242 if obj.hasFeatures():
240 for f in obj.getFeatures(): 243 for f in obj.getFeatures():
251 a = max(ymax - ymin + (2 * yMm), xmax - (xmin + 2 * xMm)) 254 a = max(ymax - ymin + (2 * yMm), xmax - (xmin + 2 * xMm))
252 yCropMin = int(max(0, .5 * (ymin + ymax - a))) 255 yCropMin = int(max(0, .5 * (ymin + ymax - a)))
253 yCropMax = int(min(height - 1, .5 * (ymin + ymax + a))) 256 yCropMax = int(min(height - 1, .5 * (ymin + ymax + a)))
254 xCropMin = int(max(0, .5 * (xmin + xmax - a))) 257 xCropMin = int(max(0, .5 * (xmin + xmax - a)))
255 xCropMax = int(min(width - 1, .5 * (xmin + xmax + a))) 258 xCropMax = int(min(width - 1, .5 * (xmin + xmax + a)))
256 if yCropMax != yCropMin and xCropMax != xCropMin and (yCropMax - yCropMin) * (xCropMax - xCropMin) > pixelThreshold: 259 if yCropMax != yCropMin and xCropMax != xCropMin and (yCropMax - yCropMin) * (xCropMax - xCropMin) > minNPixels:
257 croppedImg = img[yCropMin : yCropMax, xCropMin : xCropMax] 260 croppedImg = img[yCropMin : yCropMax, xCropMin : xCropMax]
258 else: 261 else:
259 croppedImg = [] 262 croppedImg = []
260 return croppedImg, yCropMin, yCropMax, xCropMin, xCropMax 263 return croppedImg, yCropMin, yCropMax, xCropMin, xCropMax
261 264
536 else: 539 else:
537 print(dp) 540 print(dp)
538 return None 541 return None
539 542
540 if skimageAvailable: 543 if skimageAvailable:
544 from skimage.feature import hog
545 from skimage import color, transform
546
541 def HOG(image, rescaleSize = (64, 64), orientations=9, pixelsPerCell=(8, 8), cellsPerBlock=(2, 2), visualize=False, normalize=False): 547 def HOG(image, rescaleSize = (64, 64), orientations=9, pixelsPerCell=(8, 8), cellsPerBlock=(2, 2), visualize=False, normalize=False):
542 from skimage.feature import hog
543 from skimage import color, transform
544
545 bwImg = color.rgb2gray(image) 548 bwImg = color.rgb2gray(image)
546 inputImg = transform.resize(bwImg, rescaleSize) 549 inputImg = transform.resize(bwImg, rescaleSize)
547 features = hog(inputImg, orientations, pixelsPerCell, cellsPerBlock, visualize, normalize) 550 features = hog(inputImg, orientations, pixelsPerCell, cellsPerBlock, visualize, normalize)
548 if visualize: 551 if visualize:
549 from matplotlib.pyplot import imshow, figure, subplot 552 from matplotlib.pyplot import imshow, figure, subplot
550 hogViz = features[1] 553 hogViz = features[1]
551 features = features[0] 554 features = features[0]
552 figure() 555 figure()
553 subplot(1,2,1) 556 subplot(1,2,1)
554 imshow(img) 557 imshow(inputImg)
555 subplot(1,2,2) 558 subplot(1,2,2)
556 imshow(hogViz) 559 imshow(hogViz)
557 return features 560 return float32(features)
558 561
559 def createHOGTrainingSet(imageDirectory, classLabel, rescaleSize = (64, 64), orientations=9, pixelsPerCell=(8, 8), cellsPerBlock=(2, 2), visualize=False, normalize=False): 562 def createHOGTrainingSet(imageDirectory, classLabel, rescaleSize = (64, 64), orientations=9, pixelsPerCell=(8, 8), cellsPerBlock=(2, 2), visualize=False, normalize=False):
560 from os import listdir 563 from os import listdir
561 from numpy import float32
562 from matplotlib.pyplot import imread 564 from matplotlib.pyplot import imread
563 565
564 inputData = [] 566 inputData = []
565 for filename in listdir(imageDirectory): 567 for filename in listdir(imageDirectory):
566 img = imread(imageDirectory+filename) 568 img = imread(imageDirectory+filename)