print "done"
alltrainfeatures = numpy.vstack(allbigramfeatures[:numtrain + numvali])
testfeatures = numpy.vstack(allbigramfeatures[numtrain + numvali:])
trainfeatures = alltrainfeatures[:numtrain]
valifeatures = alltrainfeatures[numtrain:]
alltrainlabels = trainlabels
valilabels = trainlabels[numtrain:]
trainlabels = trainlabels[:numtrain]
del allbigramfeatures
#CLASSIFICATION
#weightcosts = [0.1, 0.01, 0.001, 0.0001, 0.00001, 0.0]
weightcosts = [0.01, 0.001, 0.0001, 0.0]
valicosts = []
lr = logreg.Logreg(numclasses, trainfeatures.shape[1])
lr.train(trainfeatures.T,
trainlabels.T,
numsteps=100,
verbose=False,
weightcost=weightcosts[0])
lr.train_cg(trainfeatures.T,
trainlabels.T,
weightcost=weightcosts[0],
maxnumlinesearch=100)
valicosts.append(lr.zeroone(valifeatures.T, valilabels.T))
for wcost in weightcosts[1:]:
lr.train(trainfeatures.T,
trainlabels.T,
numsteps=100,
verbose=False,
pooled_features[:, (i_v * h_sections + i_h) *
num_centroids:(i_v * h_sections + i_h + 1) *
num_centroids] = mean(
feature_map[:, i_v * h_sections + i_h], 1)
if array_im.shape[0] == 1:
pooled_features = reshape(pooled_features,
(v_sections * h_sections * num_centroids))
return pooled_features.T
numclasses = 7
wc = 1e-3
lr = logreg.Logreg(numclasses, v_sections * h_sections * num_centroids)
lr.weights = loadtxt(os.path.join(params_dir, "weights_submitted.txt"))
lr.biases = loadtxt(os.path.join(params_dir, "biases_submitted.txt"))
test = []
#for j in sorted(os.listdir(target_dir)):
for j in clip_ids:
for k in sorted(os.listdir(os.path.join(target_dir, j))):
test.append(
imread(os.path.join(target_dir, j, k))[v_min:v_max, h_min:h_max])
test = asarray(test)
num_test_images = shape(test)[0]
test_features = zeros(