Annotate pictures with binary labels for given classes and keybindings. Experimental, in development.
Press space to go to the next picture, press return to go to the previous picture.
Given a file that specifies keybindings (argument BINDINGS_JSON), the program will go through the images in IMAGE_DIR
and you can switch the annotation of a class using the specified keybindings. It will save to output file --o
(default: output.csv
). With every new annotation, the annotations are saved to the output file.
It is also possible to start from an existing annotation by using the --csv
argument. Currently, the keybinding categories have to exactly match the column names of the annotation csv (it cannot be a subset at the time). The output file is different to the input --csv
argument so that it does not overwrite.
pip install git+https://github.com/jacobhepkema/annotpics.git
Usage: annotpics [OPTIONS] IMAGE_DIR BINDINGS_JSON
EXPERIMENTAL Picture annotation tool. CSV should contain 'image' column as
first column specifying image names without the directory.
Options:
--csv TEXT Location of annotation CSV file [None]
--o TEXT File to save annotations to [output.csv]
--start_i INTEGER Index of image to start with [0]
--window_w INTEGER Window width in pixels
--window_h INTEGER Window height in pixels
--resize_w INTEGER
--resize_h INTEGER
--filetype TEXT Filetype of images in IMAGE_DIR [jpg]
--version Show the version and exit.
--help Show this message and exit.
Example key binding config JSON file:
{
"bindings": {
"a": "a_category",
"b": "b_category",
"c": "c_category",
"d": "d_category",
"e": "e_category",
"f": "f_category"
}
}