It helps to test the initial mask of foreground and background.
It generates color histograms which represent foreground and background.
First segmenting the area of leaves, and then compute the area of leaves.
- Define appropriate initial mask pixel location by using
test_seg_mask.py
ref_fg = [350, 450, 1800, 1910]
ref_bg = [0, 240, 0, 2500]
- Generate color histogram by using
generate_histogram.py
Multiple reigon from multiple images could be selected.
Example of generate_histogram.bat
follows this structure.
python generate_histogram.py images\image-1550434545.jpg images\image-1550079998.jpg images\image-1550434545.jpg images\image-1550434545.jpg ^
--init_fg_masks 350 450 1800 1910 ^
--init_fg_masks 880 940 1820 1920 ^
--init_fg_masks 1450 1500 1850 1940 ^
--init_fg_masks 950 1000 1200 1250 ^
--init_bg_masks 0 240 0 2500 ^
--init_bg_masks 250 320 0 500 ^
--init_bg_masks 0 1900 1420 1750 ^
--init_bg_masks 1000 1500 1000 1100 ^
--work_dir ground_data
You will see histograms as below.
- Segment images based on the color histogram, and then optimize the segmentation mask by using
cut.py
Example of cut.bat
follows this structure.
python cut.py images\image-1550434545.jpg --histograms ground_data\histograms.npy
You will see segmentation and its area as below.
Run pip install -r requrements.txt
to install all the dependencies.
You may build and install pygco from source to run graphic cut efficiently: URL