File size: 6,841 Bytes
9b855a7 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 |
import math
import numpy as np
import cv2
def db_eval_iou(annotation, segmentation, void_pixels=None):
""" Compute region similarity as the Jaccard Index.
Arguments:
annotation (ndarray): binary annotation map.
segmentation (ndarray): binary segmentation map.
void_pixels (ndarray): optional mask with void pixels
Return:
jaccard (float): region similarity
"""
assert annotation.shape == segmentation.shape, \
f'Annotation({annotation.shape}) and segmentation:{segmentation.shape} dimensions do not match.'
annotation = annotation.astype(np.bool)
segmentation = segmentation.astype(np.bool)
if void_pixels is not None:
assert annotation.shape == void_pixels.shape, \
f'Annotation({annotation.shape}) and void pixels:{void_pixels.shape} dimensions do not match.'
void_pixels = void_pixels.astype(np.bool)
else:
void_pixels = np.zeros_like(segmentation)
# Intersection between all sets
inters = np.sum((segmentation & annotation) & np.logical_not(void_pixels), axis=(-2, -1))
union = np.sum((segmentation | annotation) & np.logical_not(void_pixels), axis=(-2, -1))
j = inters / union
if j.ndim == 0:
j = 1 if np.isclose(union, 0) else j
else:
j[np.isclose(union, 0)] = 1
return j
def db_eval_boundary(annotation, segmentation, void_pixels=None, bound_th=0.008):
assert annotation.shape == segmentation.shape
if void_pixels is not None:
assert annotation.shape == void_pixels.shape
if annotation.ndim == 3:
n_frames = annotation.shape[0]
f_res = np.zeros(n_frames)
for frame_id in range(n_frames):
void_pixels_frame = None if void_pixels is None else void_pixels[frame_id, :, :, ]
f_res[frame_id] = f_measure(segmentation[frame_id, :, :, ], annotation[frame_id, :, :], void_pixels_frame, bound_th=bound_th)
elif annotation.ndim == 2:
f_res = f_measure(segmentation, annotation, void_pixels, bound_th=bound_th)
else:
raise ValueError(f'db_eval_boundary does not support tensors with {annotation.ndim} dimensions')
return f_res
def f_measure(foreground_mask, gt_mask, void_pixels=None, bound_th=0.008):
"""
Compute mean,recall and decay from per-frame evaluation.
Calculates precision/recall for boundaries between foreground_mask and
gt_mask using morphological operators to speed it up.
Arguments:
foreground_mask (ndarray): binary segmentation image.
gt_mask (ndarray): binary annotated image.
void_pixels (ndarray): optional mask with void pixels
Returns:
F (float): boundaries F-measure
"""
assert np.atleast_3d(foreground_mask).shape[2] == 1
if void_pixels is not None:
void_pixels = void_pixels.astype(np.bool)
else:
void_pixels = np.zeros_like(foreground_mask).astype(np.bool)
bound_pix = bound_th if bound_th >= 1 else \
np.ceil(bound_th * np.linalg.norm(foreground_mask.shape))
# Get the pixel boundaries of both masks
fg_boundary = _seg2bmap(foreground_mask * np.logical_not(void_pixels))
gt_boundary = _seg2bmap(gt_mask * np.logical_not(void_pixels))
from skimage.morphology import disk
# fg_dil = binary_dilation(fg_boundary, disk(bound_pix))
fg_dil = cv2.dilate(fg_boundary.astype(np.uint8), disk(bound_pix).astype(np.uint8))
# gt_dil = binary_dilation(gt_boundary, disk(bound_pix))
gt_dil = cv2.dilate(gt_boundary.astype(np.uint8), disk(bound_pix).astype(np.uint8))
# Get the intersection
gt_match = gt_boundary * fg_dil
fg_match = fg_boundary * gt_dil
# Area of the intersection
n_fg = np.sum(fg_boundary)
n_gt = np.sum(gt_boundary)
# % Compute precision and recall
if n_fg == 0 and n_gt > 0:
precision = 1
recall = 0
elif n_fg > 0 and n_gt == 0:
precision = 0
recall = 1
elif n_fg == 0 and n_gt == 0:
precision = 1
recall = 1
else:
precision = np.sum(fg_match) / float(n_fg)
recall = np.sum(gt_match) / float(n_gt)
# Compute F measure
if precision + recall == 0:
F = 0
else:
F = 2 * precision * recall / (precision + recall)
return F
def _seg2bmap(seg, width=None, height=None):
"""
From a segmentation, compute a binary boundary map with 1 pixel wide
boundaries. The boundary pixels are offset by 1/2 pixel towards the
origin from the actual segment boundary.
Arguments:
seg : Segments labeled from 1..k.
width : Width of desired bmap <= seg.shape[1]
height : Height of desired bmap <= seg.shape[0]
Returns:
bmap (ndarray): Binary boundary map.
David Martin <dmartin@eecs.berkeley.edu>
January 2003
"""
seg = seg.astype(np.bool)
seg[seg > 0] = 1
assert np.atleast_3d(seg).shape[2] == 1
width = seg.shape[1] if width is None else width
height = seg.shape[0] if height is None else height
h, w = seg.shape[:2]
ar1 = float(width) / float(height)
ar2 = float(w) / float(h)
assert not (
width > w | height > h | abs(ar1 - ar2) > 0.01
), "Can" "t convert %dx%d seg to %dx%d bmap." % (w, h, width, height)
e = np.zeros_like(seg)
s = np.zeros_like(seg)
se = np.zeros_like(seg)
e[:, :-1] = seg[:, 1:]
s[:-1, :] = seg[1:, :]
se[:-1, :-1] = seg[1:, 1:]
b = seg ^ e | seg ^ s | seg ^ se
b[-1, :] = seg[-1, :] ^ e[-1, :]
b[:, -1] = seg[:, -1] ^ s[:, -1]
b[-1, -1] = 0
if w == width and h == height:
bmap = b
else:
bmap = np.zeros((height, width))
for x in range(w):
for y in range(h):
if b[y, x]:
j = 1 + math.floor((y - 1) + height / h)
i = 1 + math.floor((x - 1) + width / h)
bmap[j, i] = 1
return bmap
if __name__ == '__main__':
from davis2017.davis import DAVIS
from davis2017.results import Results
dataset = DAVIS(root='input_dir/ref', subset='val', sequences='aerobatics')
results = Results(root_dir='examples/osvos')
# Test timing F measure
for seq in dataset.get_sequences():
all_gt_masks, _, all_masks_id = dataset.get_all_masks(seq, True)
all_gt_masks, all_masks_id = all_gt_masks[:, 1:-1, :, :], all_masks_id[1:-1]
all_res_masks = results.read_masks(seq, all_masks_id)
f_metrics_res = np.zeros(all_gt_masks.shape[:2])
for ii in range(all_gt_masks.shape[0]):
f_metrics_res[ii, :] = db_eval_boundary(all_gt_masks[ii, ...], all_res_masks[ii, ...])
# Run using to profile code: python -m cProfile -o f_measure.prof metrics.py
# snakeviz f_measure.prof
|