def l_method(ax, X, method):
l_method = agglomerative_l_method(X, method=method)
suggest_n = len(l_method.cluster_centers_)
cmap = get_cmap(suggest_n + 1)
for label in range(suggest_n):
XX = list(map(lambda xy: xy[0],
filter(lambda xy: xy[1] == label,
zip(X, l_method.labels_))))
plot(ax, XX, c=cmap(label), edgecolors='none')
def plot(ax, sort_fn, name=''):
data = {
'acc_kmeans_3':
list(map(lambda x: x[0] / x[1], result['evaluation_kmeans_3'])),
'badness_naive':
list(map(lambda x: x['md'], result['badness_naive'])),
'badness_hierarchical_voronoid_filling': result['badness_hierarchical_voronoid_filling'],
'voronoid_sigmoid': result['voronoid_sigmoid'],
'names': result['name'],
}
# sigmoids are same for all rows
sigmoids = data['voronoid_sigmoid'][0]
# transpose the dictionary
keys = data.keys()
seq = []
for values in zip(*data.values()):
seq.append(dict(zip(keys, values)))
# sort by sort_fn
seq.sort(key=sort_fn)
# transpose it back!
def dict_to_list(d, order):
l = [d[o] for o in order]
return l
sorted_data_wo_keys = list(map(lambda d: dict_to_list(d, keys),
seq))
sorted_data = dict(zip(keys, zip(*sorted_data_wo_keys)))
# Example data
cnt = len(sorted_data['names'])
x = range(cnt)
col = sorted_data
ax.plot(x, col['acc_kmeans_3'], 'k--', color="black", label='acc c*3')
ax.plot(x, col['badness_naive'], 'k', color='grey', label='naive')
hvf = result['badness_hierarchical_voronoid_filling']
hvf_by_sigmoid = [[] for i in range(len(sigmoids))]
for each in hvf:
for i, s in enumerate(each):
hvf_by_sigmoid[i].append(s)
cmap = get_cmap(len(sigmoids) + 1)
print('dataset:', dataset)
for i, (sigmoid, each_hvf) in enumerate(zip(sigmoids, hvf_by_sigmoid)):
ax.plot(x, each_hvf, 'k', color=cmap(i), label=sigmoid)
penalty = decreasing_penalty(each_hvf)
if i == 2:
# the best now
print('penalty:', penalty)
avgs[i] += penalty
# remove y axis
ax.yaxis.set_major_formatter(plt.NullFormatter())
# scale y to [0,1]
ax.set_ylim([0, 1])
plt.sca(ax)
title = dataset.replace('_with_test', '')
plt.title(title)
# increase space between rows
plt.subplots_adjust(hspace=.5)
# rename the xticks
col_names = []
for col_name in col['names']:
if col_name.startswith('some'):
col_name = col_name.replace('some-', '')
col_name, *_ = col_name.split('-prob')
elif col_name.startswith('prob'):
col_name = col_name.replace('prob-', '')
col_names.append(col_name)
plt.xticks(range(cnt), col_names, rotation=90)
for i in range(output.shape[0]):
for j in range(output.shape[1]):
holder[i, j] = p2n[output[i, j]]
return holder
def performance(actual, predicted):
acc = accuracy_score(actual.ravel(), predicted.ravel())
prec = precision_score(actual.ravel(), predicted.ravel(), average='weighted', labels=np.unique(actual)[1:])
rec = recall_score(actual.ravel(), predicted.ravel(), average='weighted', labels=np.unique(actual)[1:])
return np.array([acc, prec, rec])
if __name__ == "__main__":
slide_cam_labelled_path = 'tiff/slide_cam_2_extra_slices/OASIS-TRT-20-'
n2id, cmap = get_cmap()
id2n = {v: k for k, v in n2id.items()}
print(len(n2id))
test_sub_IDs = range(1, 21)
fs_perf = np.zeros((len(test_sub_IDs), 3))
sc_perf = np.zeros((len(test_sub_IDs), 3))
for j, sub_ID in enumerate(test_sub_IDs):
unlabelled, man_labelled, FS_labelled = subject_volumes(sub_ID, '')
uld = unlabelled.get_data()
mld = man_labelled.get_data()
print(len(np.unique(mld)))
# fsld = FS_labelled.get_data()
