def prepare_data(k, frt, feature, master):
min_max_scaler = MinMaxScaler()
if master:
images, meta, matrix = get_full_matrix(feature, master=True)
matrix = min_max_scaler.fit_transform(matrix)
matrix, _, _ = reducer(matrix, k, frt)
# Image-Image similarity
img_img = 1 / (euclidean_distances(matrix) + 1)
np.fill_diagonal(img_img, 0)
return images, meta, img_img
l_images, l_meta, l_matrix = get_full_matrix(feature)
u_images, u_meta, u_matrix = get_full_matrix(feature, unlabelled=True)
meta = l_meta
meta.update(u_meta)
matrix = min_max_scaler.fit_transform(np.vstack((
l_matrix,
u_matrix,
)))
matrix, _, _ = reducer(matrix, k, frt)
# Image-Image similarity
img_img = 1 / (euclidean_distances(matrix) + 1)
np.fill_diagonal(img_img, 0)
return l_images + u_images, meta, img_img
def svm_feedback(relevant_images, irrelevant_images, images_to_display, query,
results):
if not irrelevant_images:
print("Please provide irrelevant images set for svm to work properly.")
return relevant_images
model = settings.SVM.CLASSIFIER.MODEL
k = settings.SVM.CLASSIFIER.K
frt = settings.SVM.CLASSIFIER.FRT
images_rel, data_matrix_rel = get_all_vectors(
model, f={'path': {
'$in': relevant_images
}}, master_db=True)
images_irel, data_matrix_irel = get_all_vectors(
model, f={'path': {
'$in': irrelevant_images
}}, master_db=True)
images_test, test_vector = get_all_vectors(
model, f={'path': {
'$in': results + [query]
}}, master_db=True)
labelled_vectors, _, _, unlabelled_vectors = reducer(
np.vstack((data_matrix_rel, data_matrix_irel)),
k,
frt,
query_vector=test_vector)
rel_class = np.array([1] * len(data_matrix_rel))
irel_class = np.array([-1] * len(data_matrix_irel))
x_train = labelled_vectors
x_train = np.array(x_train) * 2
y_train = np.concatenate((rel_class, irel_class))
svclassifier = SVM()
svclassifier.fit(np.array(x_train), np.array(y_train))
unlabelled_vectors = np.array(unlabelled_vectors) * 2
y_pred = svclassifier.predict(unlabelled_vectors)
c = 0
dic = {}
for y in y_pred:
if y == 1:
dic[images_test[c]] = unlabelled_vectors[c]
c += 1
length_dict = {}
for key in dic.keys():
length_dict[key] = np.dot(dic[key], svclassifier.w)
sorted_dict = sorted(length_dict.items(), key=lambda x: x[1], reverse=True)
list_img = []
c = 0
for key, j in sorted_dict:
if c < images_to_display - len(irrelevant_images):
list_img.append(key)
c += 1
else:
break
return (list_img + irrelevant_images)
def prepare_data(k, frt, feature, paths=None):
min_max_scaler = MinMaxScaler()
if paths:
images, meta, matrix = get_data_matrix(feature,
f={'path': {
'$in': paths
}})
else:
images, meta, matrix = get_data_matrix(feature)
matrix = min_max_scaler.fit_transform(matrix)
matrix, _, _ = reducer(matrix, k, frt)
return images, meta, matrix
def run_svm(evaluate, model='lbp', k=30, frt='pca'):
train_data, train_labels = build_labelled(model)
if evaluate:
test_data, test_labels, test_paths = build_unlabelled(model)
else:
test_paths, test_data = get_all_vectors(model, f={}, unlabelled_db=True)
labelled_vectors, _, _, unlabelled_vectors = reducer(train_data, k, frt, query_vector=test_data)
labelled_vectors *= 2
unlabelled_vectors *= 2
svclassifier = SVM()
svclassifier.fit(labelled_vectors, train_labels)
y_pred = svclassifier.predict(unlabelled_vectors)
if evaluate:
print(classification_report(test_labels,y_pred))
return test_paths, y_pred
def prepare_data(cls, feature, k_latent_semantics, frt_technique,
ignore_metadata):
# Get the images from the folders specified in config
# We expect the vectors to be build for the features for both the labelled
# and unlabelled data.
u_images, u_meta, u_matrix = cls.get_data_matrix(
feature, unlabelled=True, ignore_metadata=ignore_metadata)
l_images, l_meta, l_matrix = cls.get_data_matrix(
feature, ignore_metadata=ignore_metadata)
# Reduce the labeled and unlabeled matrix together
old_matrix = np.vstack((
l_matrix,
u_matrix,
))
matrix, _, _ = reducer(old_matrix, k_latent_semantics, frt_technique)
r_l_matrix = matrix[:len(l_images)]
r_u_matrix = matrix[len(l_images):]
return l_images, u_images, l_meta, u_meta, r_l_matrix, r_u_matrix
def feedback_probab(relevant, irrelevant, t, query, prev_results):
if not relevant:
print("Probabilistic model requires relevant images for re-ordering.")
return prev_results
img_all, img_all_vec = get_all_vectors(
model, f={'path': {
'$in': prev_results + [query]
}}, master_db=True)
#f={'path': {'$nin': relevant}}
img_all_vec_red, _, __ = reducer(img_all_vec, k, frt)
img_all_vec_red = scale(img_all_vec_red, 0, 1)
dict_all_red = {}
for i in range(len(img_all)):
name = img_all[i]
dict_all_red[name] = img_all_vec_red[i]
img_rel_vec_red = []
for name in relevant:
img_rel_vec_red.append(dict_all_red[name])
img_rel_vec_red = np.array(img_rel_vec_red)
img_all_vec_red = makeArrayBinary(img_all_vec_red,
img_all_vec_red.shape[0],
img_all_vec_red.shape[1])
img_rel_vec_red = makeArrayBinary(img_rel_vec_red,
img_rel_vec_red.shape[0],
img_rel_vec_red.shape[1])
R = img_rel_vec_red.shape[0]
N = len(img_all)
p_list = []
for j in range(k):
r = 0
for i in range(R):
if img_rel_vec_red[i][j] == 1:
r += 1
p_list.append((r + 0.5) / (R + 1))
n_list = []
for j in range(k):
n = 0
for i in range(N):
if img_all_vec_red[i][j] == 1:
n += 1
n_list.append(n)
for i in range(k):
n_list[i] = (n_list[i] - p_list[i] + 0.5) / (N - R + 1)
log_list = []
for i in range(k):
num = (p_list[i] * (1 - n_list[i])) / (n_list[i] * (1 - p_list[i]))
if num > 0:
log_list.append(math.log(num, 2))
log_list = np.array(log_list)
new_result = []
for name in dict_all_red.keys():
sim = np.dot(dict_all_red[name], log_list)
new_result.append((name, sim))
new_result = sorted(new_result, key=lambda x: x[1], reverse=True)
final = []
for i in range(t):
final.append(new_result[i][0])
return final
# Only 1 input to be taken, that is k (latent semantics), number of features to be extracted
k_each = args.Features
# On extensive testing, the best feature extraction model was found out to be SIFT
model = settings.TASK1_CONFIG.MODEL
# On extensive testing, the best feature reduction technique was founf out to be PCA
feature = settings.TASK1_CONFIG.FRT
# Generating the vectors for Dorsal Labelled, Palmar labelled and the test vectors
# Also fetching the labels of unlabelled images so as to check accuracy later
dorsal_vectors, palmar_vectors, test_data, test_data_paths = generate_vec()
# Applying PCA to Dorsal Images and fetching the 'k' latent semantics
reduced_dorsal_vectors, _, _, _, dorsal_pca = reducer(dorsal_vectors,k_each,feature,get_scaler_model=True)
dorsal_variance_ratio = dorsal_pca.explained_variance_ratio_
print("Computed ",k_each," Latent Semantics for Dorsal")
# Applying PCA to Palmar Images and fetching the 'k' latent semantics
reduced_palmar_vectors, _, _, _, palmar_pca = reducer(palmar_vectors,k_each,feature,get_scaler_model=True)
palmar_variance_ratio = palmar_pca.explained_variance_ratio_
print("Computed ",k_each," Latent Semantics for Palmar")
# Applying PCA to Test Images and fetching the 'k' latent semantics
reduced_test_data, _, _, _, test_pca = reducer(test_data,k_each,feature,get_scaler_model=True)
test_variance_ratio = test_pca.explained_variance_ratio_
# Initiate List that will store the total dorsal dot product scores for each test image
dorsal = []
images, feature_space = utils.get_all_vectors(
settings.PPR.CLASSIFIER.FEATURE)
feature_space = min_max_scaler.fit_transform(feature_space)
meta = utils.get_metadata()
meta = {m['path']: m for m in meta}
"""
u_images, u_feature_space = utils.get_all_vectors(
settings.PPR.CLASSIFIER.FEATURE, unlabelled_db=True)
u_feature_space = min_max_scaler.fit_transform(u_feature_space)
matrix = np.vstack((
feature_space,
u_feature_space,
))
matrix, eigen_values, latent_vs_old = reducer(
matrix,
"""
matrix, eigen_values, latent_vs_old = reducer(feature_space,
settings.PPR.CLASSIFIER.K,
settings.PPR.CLASSIFIER.FRT)
dm = helper.build_matrix_with_labels(matrix, images, meta)
"""
dm = helper.build_labelled_matrix(matrix, images + u_images,
'aspectOfHand')
"""
evaluate(dm)
sub_meta[meta[img]['id']] = meta[img]
# sub id to order in matrix
sub_to_idx = {sub: idx for idx, sub in enumerate(subs)}
# index to sub id
idx_to_sub = [0] * len(sub_to_idx)
for sub in sub_to_idx:
idx_to_sub[sub_to_idx[sub]] = sub
# A subject subject similarity index
sub_sub = np.zeros((len(subs), len(subs),))
for sub1 in sub_to_idx:
for sub2 in sub_to_idx:
sub_sub[sub_to_idx[sub1], sub_to_idx[sub2]] = img_img[subs[sub1],:].take(subs[sub2], axis=1).mean()
w, _, h = reducer(sub_sub, args.k_latent_semantics, "nmf")
# Print term weigth pairs
get_term_weight_pairs(w, "task7_{}.csv".format(args.k_latent_semantics))
sub_weight = [
sorted([("z{}".format(idx), weight,) for idx, weight in enumerate(row)], key=lambda x: x[1])
for row in w
]
output.write_to_file("visualize_task7.html",
"task7-{}.html".format(args.k_latent_semantics),
vectors=sub_weight,
subs=subs,
idx_to_sub=idx_to_sub,
images=images,
sub_meta=sub_meta,
img_meta = []
try:
for m in meta:
images.append(m['path'])
img_meta.append([
m["age"], mapping[m["gender"]], mapping[m["skinColor"]],
mapping[m["accessories"]], m["nailPolish"],
mapping[m["aspectOfHand"].split()[0]],
mapping[m["aspectOfHand"].split()[1]], m["irregularities"]
])
except KeyError:
raise Exception("Invalid metadata detected")
vectors, eigen_values, latent_vs_old = reducer(img_meta,
args.k_latent_semantics,
"nmf")
get_term_weight_pairs(vectors,
"task8_{}.csv".format(args.k_latent_semantics))
get_term_weight_pairs(latent_vs_old,
"task8_{}.csv".format(args.k_latent_semantics))
# Extra Credit
# image path with a vector in the latent semantic space
data_z = zip(images, vectors)
# image path for each latenet semantic in h
feature_z = [(idx, images[np.argmax(np.dot(img_meta, i))])
for idx, i in enumerate(latent_vs_old)]
output.write_to_file("visualize_data_z.html",
