parser.add_argument("--dcompress", help="Number of dimensions to compress " "features to.", type=int, default=None) parser.add_argument("--npatches", help="Number of image patches to use to learn" " dictionary.", type=int, default=200000) args = parser.parse_args() # Make a list of images filelist = glob.glob(os.path.join(args.imagedir, '*.' + args.extension)) nimages = len(filelist) print "Found {0} images.".format(nimages) if nimages == 0: print "Quiting..." sys.exit(1) # Train a dictionary desc = ScSPM(dsize=args.nbases, compress_dim=args.dcompress) desc.learn_dictionary(filelist, npatches=args.npatches, niter=5000) # Save the dictionary with open(args.dicname, 'wb') as f: pk.dump(desc, f, protocol=2) print "Done! Dictionary object saved to {0}.".format(args.dicname)
from imdescrip.extractor import extract_smp from imdescrip.descriptors.ScSPM import ScSPM # Make a list of images imgdir = '/home/dsteinberg/Datasets/Tas2008_12/Images/' savdir = '/home/dsteinberg/Datasets/Tas2008_12/Images/imdesc/' filelist = glob.glob(imgdir + '*.png') #imgdir = '/home/dsteinberg/Datasets/outdoor_scenes/Images/' #savdir = '/home/dsteinberg/Datasets/outdoor_scenes/Images/imdesc/' #filelist += glob.glob(imgdir + '*.jpg') #imgdir = '/home/dsteinberg/Datasets/Caltech101_sub/' #filelist += glob.glob(imgdir + '*.jpg') #print len(filelist) # Train a dictionary desc = ScSPM(dsize=512, compress_dim=None) desc.learn_dictionary(filelist, npatches=200000, niter=5000) # Save the dictionary with open('ScSPM.p', 'wb') as f: pk.dump(desc, f, protocol=2) # OR Load a pre-learned dictionary #with open('ScSPM.p', 'rb') as f: #desc = pk.load(f) extract_smp(filelist, savdir, desc, verbose=True)
import cPickle as pk from imdescrip.extractor import extract_smp from imdescrip.descriptors.ScSPM import ScSPM # Make a list of images imgdir = '/home/dsteinberg/Datasets/Tas2008_12/Images/' savdir = '/home/dsteinberg/Datasets/Tas2008_12/Images/imdesc/' filelist = glob.glob(imgdir + '*.png') #imgdir = '/home/dsteinberg/Datasets/outdoor_scenes/Images/' #savdir = '/home/dsteinberg/Datasets/outdoor_scenes/Images/imdesc/' #filelist += glob.glob(imgdir + '*.jpg') #imgdir = '/home/dsteinberg/Datasets/Caltech101_sub/' #filelist += glob.glob(imgdir + '*.jpg') #print len(filelist) # Train a dictionary desc = ScSPM(dsize=512, compress_dim=None) desc.learn_dictionary(filelist, npatches=200000, niter=5000) # Save the dictionary with open('ScSPM.p', 'wb') as f: pk.dump(desc, f, protocol=2) # OR Load a pre-learned dictionary #with open('ScSPM.p', 'rb') as f: #desc = pk.load(f) extract_smp(filelist, savdir, desc, verbose=True)
parser.add_argument("extension", help="Image file extension (eg. 'png').") parser.add_argument("--dicname", help="Name and path of dictionary file to " "save.", default="ScSPM.p") parser.add_argument("--nbases", help="Number of dictionary bases.", type=int, default=512) parser.add_argument("--dcompress", help="Number of dimensions to compress " "features to.", type=int, default=None) parser.add_argument("--npatches", help="Number of image patches to use to learn" " dictionary.", type=int, default=200000) args = parser.parse_args() # Make a list of images filelist = glob.glob(os.path.join(args.imagedir, '*.' + args.extension)) nimages = len(filelist) print "Found {0} images.".format(nimages) if nimages == 0: print "Quiting..." sys.exit(1) # Train a dictionary desc = ScSPM(dsize=args.nbases, compress_dim=args.dcompress) desc.learn_dictionary(filelist, npatches=args.npatches, niter=5000) # Save the dictionary with open(args.dicname, 'wb') as f: pk.dump(desc, f, protocol=2) print "Done! Dictionary object saved to {0}.".format(args.dicname)