I've built scikit-image 0.14 from source: https://github.com/scikit-image/scikit-image/blob/master/README.md to test some of the new features, but I think 0.13 (the stable release) should also still work for now.

Typical usage: python3 superpixels.py -i path/to/image The program will automaticall dump the index matrix to index_matrix.csv and the relevant matching data (index, x, y, center) to matching_data.csv

Typical usage: python3 superpixels.py -i path/to/image -d datafile.txt -m (This generates a mask for each keyframe specified in datafile.txt)

• -d datafile.txt allows the specification of target segments for matching, using Zach's format (TODO: Add here).

• If the -m flag is given, these will be used to generate a mask for each keyframe of the image.

• The -di flag may be used to display the image(s) using matplotlib; by default each will be output to a file output_x.txt instead

• The -s size flag may be used to manually specify the maximum distance between pixels in a segment (larger values will lead to generally larger superpixels).

• The -q qr_file.txt flag be may be used to specify a file containing two integer coordinates of a pixel in the position marker of a QR code, so that position can be matched relative to the QR code.

Run the matching gui with python3 gui.py

This feature can be used for development to compare the segmentations of similar images, though using the gui is probably easier.

To match segments between two images: python3 superpixels.py -i path/to/image -i2 /path/to/other/image -d datafile.txt

where datafile.txt contains the pixels of the qr codes in each image, along with pixels in image 1 whose corresponding segments should be matched with segments in image2: qrxpixel1 qrypixel1 qexpixel2 qrypixel2 point1x point1y point2x point2y ...

virtualenv VirtualEnv

source VirtualEnv/bin/activate

pip3 install -U scikit-image

cd segmentation

Segmentation with csv as output:

python3 superpixels.py -i img_author.png

Matching with masks as output:

python3 superpixels.py -i img_user.png -d qr_author.txt -m