Image statistics
Requirements#
Requires the image-dataset-converter-imgstats library.
Download the 17 flowers object detection VOC XML dataset and extract it.
Plugins#
Label distribution#
Using the label-dist
writer, you can output a distribution of the labels within
a dataset as follows:
idc-convert \
-l INFO \
from-voc-od \
-l INFO \
-i "./voc/*.xml" \
label-dist \
-l INFO
This will output the following:
Bluebell: 28
Buttercup: 54
ColtsFoot: 55
Crocus: 50
Daffodil: 71
Daisy: 57
Dandelion: 43
Fritillary: 65
Iris: 77
LilyValley: 17
Pansy: 56
Snowdrop: 50
Sunflower: 71
Tigerlily: 50
Tulip: 41
Windflower: 63
NB: You can also generate CSV or JSON output.
Area histogram#
Knowing the distribution of the extends of your annotations can be quite useful, e.g.,
for determining whether down-sampling can be an option. The area-histogram
writer
generates histograms for all objects combined and one for each category:
idc-convert \
-l INFO \
from-voc-od \
-l INFO \
-i "./voc/*.xml" \
area-histogram \
-l INFO
The output for the 17 flowers dataset looks like this:
ALL:
+1.26e+04 - +3.29e+04 [ 14] █████▎
+3.29e+04 - +5.33e+04 [ 30] ███████████▍
+5.33e+04 - +7.36e+04 [ 36] █████████████▋
+7.36e+04 - +9.40e+04 [ 54] ████████████████████▍
+9.40e+04 - +1.14e+05 [ 71] ██████████████████████████▊
+1.14e+05 - +1.35e+05 [ 71] ██████████████████████████▊
+1.35e+05 - +1.55e+05 [ 95] ███████████████████████████████████▉
+1.55e+05 - +1.75e+05 [104] ███████████████████████████████████████▎
+1.75e+05 - +1.96e+05 [ 94] ███████████████████████████████████▌
+1.96e+05 - +2.16e+05 [106] ████████████████████████████████████████
+2.16e+05 - +2.37e+05 [ 77] █████████████████████████████
+2.37e+05 - +2.57e+05 [ 48] ██████████████████▏
+2.57e+05 - +2.77e+05 [ 14] █████▎
+2.77e+05 - +2.98e+05 [ 10] ███▊
+2.98e+05 - +3.18e+05 [ 9] ███▍
+3.18e+05 - +3.38e+05 [ 6] ██▎
+3.38e+05 - +3.59e+05 [ 2] ▊
+3.59e+05 - +3.79e+05 [ 5] █▉
+3.79e+05 - +3.99e+05 [ 1] ▍
+3.99e+05 - +4.20e+05 [ 1] ▍
Bluebell:
+7.68e+04 - +9.18e+04 [1] ████████
+9.18e+04 - +1.07e+05 [1] ████████
+1.07e+05 - +1.22e+05 [2] ████████████████
+1.22e+05 - +1.37e+05 [0]
+1.37e+05 - +1.51e+05 [0]
+1.51e+05 - +1.66e+05 [2] ████████████████
+1.66e+05 - +1.81e+05 [2] ████████████████
+1.81e+05 - +1.96e+05 [0]
+1.96e+05 - +2.11e+05 [5] ████████████████████████████████████████
+2.11e+05 - +2.26e+05 [1] ████████
+2.26e+05 - +2.41e+05 [4] ████████████████████████████████
+2.41e+05 - +2.56e+05 [3] ████████████████████████
+2.56e+05 - +2.71e+05 [1] ████████
+2.71e+05 - +2.86e+05 [0]
+2.86e+05 - +3.01e+05 [2] ████████████████
+3.01e+05 - +3.16e+05 [0]
+3.16e+05 - +3.31e+05 [1] ████████
+3.31e+05 - +3.46e+05 [0]
+3.46e+05 - +3.61e+05 [0]
+3.61e+05 - +3.76e+05 [3] ████████████████████████
Buttercup:
...
NB: You can also generate CSV or JSON output.