Mercurial > ift6266
comparison data_generation/pipeline/filter_nist.py @ 626:75dbbe409578
Added code for deep mlp, experiment code to go along with it. Also added code I used to filter the P07 / PNIST07 datasets to keep only digits.
| author | fsavard |
|---|---|
| date | Wed, 16 Mar 2011 13:43:32 -0400 |
| parents | |
| children |
comparison
equal
deleted
inserted
replaced
| 625:128bc92897f2 | 626:75dbbe409578 |
|---|---|
| 1 import numpy | |
| 2 from pylearn.io import filetensor as ft | |
| 3 from ift6266 import datasets | |
| 4 from ift6266.datasets.ftfile import FTDataSet | |
| 5 | |
| 6 dataset_str = 'P07_' # NISTP # 'P07safe_' | |
| 7 | |
| 8 #base_path = '/data/lisatmp/ift6266h10/data/'+dataset_str | |
| 9 #base_output_path = '/data/lisatmp/ift6266h10/data/transformed_digits/'+dataset_str+'train' | |
| 10 | |
| 11 base_path = '/data/lisa/data/ift6266h10/data/'+dataset_str | |
| 12 base_output_path = '/data/lisatmp/ift6266h10/data/transformed_digits/'+dataset_str+'train' | |
| 13 | |
| 14 for fileno in range(100): | |
| 15 print "Processing file no ", fileno | |
| 16 | |
| 17 output_data_file = base_output_path+str(fileno)+'_data.ft' | |
| 18 output_labels_file = base_output_path+str(fileno)+'_labels.ft' | |
| 19 | |
| 20 print "Reading from ",base_path+'train'+str(fileno)+'_data.ft' | |
| 21 | |
| 22 dataset = lambda maxsize=None, min_file=0, max_file=100: \ | |
| 23 FTDataSet(train_data = [base_path+'train'+str(fileno)+'_data.ft'], | |
| 24 train_lbl = [base_path+'train'+str(fileno)+'_labels.ft'], | |
| 25 test_data = [base_path+'_test_data.ft'], | |
| 26 test_lbl = [base_path+'_test_labels.ft'], | |
| 27 valid_data = [base_path+'_valid_data.ft'], | |
| 28 valid_lbl = [base_path+'_valid_labels.ft']) | |
| 29 # no conversion or scaling... keep data as is | |
| 30 #indtype=theano.config.floatX, inscale=255., maxsize=maxsize) | |
| 31 | |
| 32 ds = dataset() | |
| 33 | |
| 34 all_x = [] | |
| 35 all_y = [] | |
| 36 | |
| 37 all_count = 0 | |
| 38 | |
| 39 for mb_x,mb_y in ds.train(1): | |
| 40 if mb_y[0] <= 9: | |
| 41 all_x.append(mb_x[0]) | |
| 42 all_y.append(mb_y[0]) | |
| 43 | |
| 44 if (all_count+1) % 100000 == 0: | |
| 45 print "Done next 100k" | |
| 46 | |
| 47 all_count += 1 | |
| 48 | |
| 49 # data is stored as uint8 on 0-255 | |
| 50 merged_x = numpy.asarray(all_x, dtype=numpy.uint8) | |
| 51 merged_y = numpy.asarray(all_y, dtype=numpy.int32) | |
| 52 | |
| 53 print "Kept", len(all_x), "(shape ", merged_x.shape, ") examples from", all_count | |
| 54 | |
| 55 f = open(output_data_file, 'wb') | |
| 56 ft.write(f, merged_x) | |
| 57 f.close() | |
| 58 | |
| 59 f = open(output_labels_file, 'wb') | |
| 60 ft.write(f, merged_y) | |
| 61 f.close() | |
| 62 |