#import dataset as ds
#import learning_dataset
from labeling.exps import results_dirs as rd
from PIL import Image, ImageFont, ImageDraw
import glob
"""
Makes plots self
"""
def gray2rgb(im):
return (color.gray2rgb(im) * 255).astype(np.uint8)
file_out = os.path.join(rd.root_dir, 'plots_results')
placehold = utls.imread(os.path.join(file_out, 'placeholder.png'))
n_sets_per_type = 2
dfs = []
# Self-learning
ims = []
ksp = []
vilar = []
g2s = []
mic17 = []
wtp = []
#for key in rd.res_dirs_dict_ksp.keys(): # Types
for key in rd.types: # Types
# Make images/gts/gaze-point
confs = rd.confs_dict_ksp[key][dset]
conf = confs[0]
dataset = ld.LearningDataset(conf)
gt = dataset.gt
ksp_mat = np.load(
os.path.join(rd.root_dir, rd.res_dirs_dict_ksp[key][dset][gset],
'results.npz'))['ksp_scores_mat']
for f in range(len(conf.frameFileNames)):
cont_gt = segmentation.find_boundaries(gt[..., f], mode='thick')
idx_cont_gt = np.where(cont_gt)
im = utls.imread(conf.frameFileNames[f])
im[idx_cont_gt[0], idx_cont_gt[1], :] = (255, 0, 0)
myGaze_fg = utls.readCsv(
os.path.join(conf.root_path, conf.ds_dir, conf.locs_dir,
conf.csvFileName_fg))
im = csv.draw2DPoint(myGaze_fg, f, im, radius=7)
ksp_ = (cmap(ksp_mat[..., f] * 255)[..., 0:3] * 255).astype(
np.uint8)
im = np.concatenate((im, ksp_), axis=0)
ims.append(im)
# Save frames
if (not os.path.exists(path_out)):
for dir_ in rd.res_dirs_dict_mic17[key]:
print('Scoring:')
print(dir_)
path_ = os.path.join(rd.root_dir, dir_)
# Get config
conf = cfg.load_and_convert(os.path.join(path_, 'cfg.yml'))
conf.root_path = rd.root_dir
conf.dataOutDir = os.path.join(rd.root_dir, conf.dataOutDir)
l_dataset = learning_dataset.LearningDataset(conf, pos_thr=0.5)
gt = l_dataset.gt
res_files = sorted(glob.glob(os.path.join(path_, '*.png')))
preds = np.asarray([utls.imread(f) for f in res_files])
preds = np.mean(preds, axis=3)
preds = preds.transpose((1, 2, 0))
pr, rc, _ = precision_recall_curve(gt.ravel(), preds.ravel())
all_f1s = 2 * (pr * rc) / (pr + rc)
max_f1 = np.nanmax(all_f1s)
max_pr = pr[np.argmax(all_f1s)]
max_rc = rc[np.argmax(all_f1s)]
file_out = os.path.join(path_, 'scores.csv')
C = pd.Index(["F1", "PR", "RC"], name="columns")
I = pd.Index(['EEL'], name="Methods")
data = np.asarray([max_f1, max_pr, max_rc]).reshape(1, 3)
df = pd.DataFrame(data=data, index=I, columns=C)
print('Saving F1 score')
res = np.load(os.path.join(rd.root_dir, dir_, 'results.npz'))
list_paths_back = res['list_paths_back']
list_paths_forw = res['list_paths_for']
color_sps = (0, 0, 255)
ims = []
ksp = []
# Load config
dataset = learning_dataset.LearningDataset(conf)
gt = dataset.gt
# Image
ims = [utls.imread(f) for f in conf.frameFileNames]
cont_gts = [segmentation.find_boundaries(gt[..., i])
for i in range(gt.shape[-1])]
idx_cont_gts = [np.where(g) for g in cont_gts]
path_csv = os.path.join(conf.root_path,
conf.ds_dir,
conf.locs_dir,
conf.csvFileName_fg)
labels = dataset.get_labels()
dataset.load_labels_contours_if_not_exist()
label_conts = dataset.labelContourMask
locs = utls.readCsv(path_csv)
for i in range(len(ims)):
ims[i][idx_cont_gts[i][0], idx_cont_gts[i][1], :] = (255, 0, 0)
ksp = np.load(file_ksp)['ksp_scores_mat']
ksp = [
cmap((ksp[..., i] * 255).astype(np.uint8))[..., 0:3]
for i in range(ksp.shape[-1])
]
with progressbar.ProgressBar(maxval=len(conf.frameFileNames)) as bar:
for f, i in zip(conf.frameFileNames,
range(len(conf.frameFileNames))):
bar.update(i)
# Image
cont_gt = segmentation.find_boundaries(gt[..., i],
mode='thick')
idx_cont_gt = np.where(cont_gt)
im = utls.imread(f)
im[idx_cont_gt[0], idx_cont_gt[1], :] = (255, 0, 0)
locs2d = csv.readCsv(
os.path.join(conf.root_path, conf.ds_dir, conf.locs_dir,
conf.csvFileName_fg))
im = csv.draw2DPoint(locs2d, i, im, radius=7)
im = im / 255
im = np.concatenate((im, ksp[i].copy()), axis=1)
im = (im * 255).astype(np.uint8)
io.imsave(
os.path.join(dir_out, 'frame_{:04d}.png'.format(i + 1)),
im)