A configurable batch script for common image operations (crop, affine transformation, split, combination, comparison, etc.).
With this workflow script, image operations can be easily & automatically performed following the guidance of which has been configured in a *.yml file, an example is as below:
(op1: split images; op2: combine them back; op3: evaluate their similarity)
opencv-python
PyYAML
numpy
scikit-image (for evaluation metrics)
- Put your image files in a structure like this:
. ├── raw │ ├── 1.png │ ├── 2.png │ ├── 3.png │ ├── 4.png │ └── ... ├── compare # only if comparison is needed │ ├── 1.png │ ├── 2.png │ ├── 3.png │ ├── 4.png │ └── ... └─── result
-
Copy one of the configure templates in 'configs' directory (e.g.
configs/workflow.yml) toconfigs/my_workflow.yml(leaving the original as a backup). -
Modify
configs/my_workflow.ymlas you need. Beware thatinput,compare,outputfields should be set to corresponding directory paths.
input: 'raw' compare: 'compare' output: 'result'
- Run
python workflow.py configs/my_config.yml
usage: workflow.py [-h] [--input INPUT] [--output OUTPUT] [--compare COMPARE]
[--yes]
[--mode {default,1_to_1,1_to_n,n_to_1,2_to_0,n_to_0}]
ymlpath
usage: python workflow.py [config].yml --input input_dir
positional arguments:
ymlpath
optional arguments:
-h, --help show this help message and exit
--input INPUT, -i INPUT
input dir (can set in configure file)
--output OUTPUT, -o OUTPUT
output dir
--compare COMPARE, -c COMPARE
compare dir
--yes, -y ignore confirmations.
--mode {default,1_to_1,1_to_n,n_to_1,2_to_0,n_to_0}
set to `default` or `{x}_to_{y}`. x: num of images
handled once, y: if n, a folder will be created for
each input image.
The following example shows the configure file to split images in folder val into 2 ×ばつ 2 tiles (saved in split folder) and then combine them back (saved in combine folder). Afterwards, our op3 will check if the inputs and combined results are exactly the same. (by ssim)
workflow: - op1: __meta__: input: 'val' # input directory output: 'split' recursively: False # recursively = False: not include sub directory split: tiles: w: 2 h: 2 - op2: __meta__: input: 'split' output: 'combine' recursively: True # recursively = True: include sub directory combine: one_folder_in_axis: 'xy' # one folder in one output image priority_axis: 'x' # row-major tiles: w: 2 # num_a_row h: 2 # rows(folders) - op3: __meta__: input: 'val' compare: 'combine' evaluate: - f1 - f2 - ssim
The running result snapshot is shown in What Can it Do part.
A more comprehensive configure template is available here.