def eval_landmarks(ref_landmark_path, output_path, cnt):
ref_lm = landmarks.read_csv(
ref_landmark_path + "ref_landmarks.csv",
False) #csv_2_np.read_csv(out1_path + "ref_landmarks.csv", False)
out_succ_freq = []
out_succ_means = []
out_means = []
out_stddevs = []
out_dists = []
out_dist_summary = []
for j in xrange(len(param_comb)):
dists = np.zeros(cnt)
for i in xrange(cnt):
tra_lm = landmarks.read_csv(output_path +
"registered_landmarks_%d_%d.csv" %
(j + 1, i + 1))
dists[i] = landmarks.mean_euclidean(ref_lm, tra_lm)
succ_freq = np.count_nonzero(
np.where(dists <= 1.0)) / np.float(cnt)
succ_means = np.mean(dists[np.where(dists <= 1.0)])
out_succ_freq.append(succ_freq)
out_succ_means.append(succ_means)
out_means.append(np.mean(dists))
out_stddevs.append(np.std(dists))
out_dists.append(dists)
out_dist_summary.append(dt.make_distribution(dists))
#np.sort(dists)
return (out_succ_freq, out_succ_means, out_means, out_stddevs,
out_dists, out_dist_summary)
def eval_landmarks(ref_landmark_path, output_path, cnt):
ref_lm = landmarks.read_csv(
ref_landmark_path + "ref_landmarks.csv",
False) #csv_2_np.read_csv(out1_path + "ref_landmarks.csv", False)
dists = np.zeros(cnt)
for i in xrange(cnt):
tra_lm = landmarks.read_csv(output_path +
"registered_landmarks_%d.csv" %
(i + 1))
dists[i] = landmarks.mean_euclidean(ref_lm, tra_lm)
#np.sort(dists)
return (np.mean(dists), np.std(dists), dists,
dt.make_distribution(dists))