Autocrop

A opencv based python script that used to crop out the 96-well colonies on a petri dish.

Usage

Use python AUTOCROP --help to see more details.

positional arguments:
  image                 Image that contains colony plates.

options:
  -h, --help            show this help message and exit
  -o OUTDIR, --outdir OUTDIR
                        Output directory.
  -n N_PLATES, --n_plates N_PLATES
                        Number of plates on the petri dish. (Default=3)
  -p {H,M,L}, --param {H,M,L}
                        Force to use parameters for high/medium/low growing plate.
  -q, --quiet           Quiet mode to suppress msg.
  -v, --verbose         Verbose mode for debug.

Let's process the image stored in original_images/3_1_30.jpg and output to cropped_images.

python AUTOCROP original_images/3_1_30.jpg -o cropped_images

By default, AUTOCROP will automatically find a proper parameters for high/medium/low growing plates. This image was detected to be a medium growing plate.

2023-09-26 14:59:03,719 autocrop INFO [3_1_30.jpg] Start processing original_images/3_1_30.jpg
2023-09-26 14:59:05,401 autocrop INFO [3_1_30.jpg] Use "M" parameter set
2023-09-26 14:59:05,517 autocrop INFO [3_1_30.jpg] Write adjusted image to output/3_1_30.jpg
2023-09-26 14:59:06,028 autocrop INFO [3_1_30.jpg] Write plate1 image to output/3_1_30_1.jpg
2023-09-26 14:59:06,045 autocrop INFO [3_1_30.jpg] Write plate2 image to output/3_1_30_2.jpg
2023-09-26 14:59:06,061 autocrop INFO [3_1_30.jpg] Write plate3 image to output/3_1_30_3.jpg
2023-09-26 14:59:06,064 autocrop INFO [3_1_30.jpg] Write marker image to output/3_1_30_M.jpg

However, the parameter auto-detection sometime will failed.

python AUTOCROP original_images/7_2_low_N.jpg -o cropped_images

2023-09-26 15:00:49,116 autocrop INFO [7_2_low_N.jpg] Start processing original_images/7_2_low_N.jpg
2023-09-26 15:00:51,092 autocrop INFO [7_2_low_N.jpg] Use "H" parameter set
2023-09-26 15:00:51,207 autocrop INFO [7_2_low_N.jpg] Write adjusted image to output/7_2_low_N.jpg
2023-09-26 15:00:51,285 autocrop WARNING [7_2_low_N.jpg] Cannot found proper dilation parameter. Aborted.

In this case, you can force to use a particular parameter set by specifying --param/-p [H/M/L]. E.g.

python AUTOCROP original_images/7_2_low_N.jpg -o cropped_images -p M

Sometimes the issue would be resolved.

2023-09-26 15:01:46,532 autocrop INFO [7_2_low_N.jpg] Start processing original_images/7_2_low_N.jpg
2023-09-26 15:01:48,468 autocrop INFO [7_2_low_N.jpg] Use "M" parameter set
2023-09-26 15:01:48,583 autocrop INFO [7_2_low_N.jpg] Write adjusted image to output/7_2_low_N.jpg
2023-09-26 15:01:48,941 autocrop INFO [7_2_low_N.jpg] Write plate1 image to output/7_2_low_N_1.jpg
2023-09-26 15:01:48,956 autocrop INFO [7_2_low_N.jpg] Write plate2 image to output/7_2_low_N_2.jpg
2023-09-26 15:01:48,970 autocrop INFO [7_2_low_N.jpg] Write plate3 image to output/7_2_low_N_3.jpg
2023-09-26 15:01:48,977 autocrop INFO [7_2_low_N.jpg] Write marker image to output/7_2_low_N_M.jpg

Experimental feature

Now you can adjust the -n/--n_plates if the petri dish is expected to contain more than 3 plates. E.g. 4 plates are expected to be found on the petri dish:

python AUTOCROP original_images/7_2_low_N.jpg -o cropped_images -n 4

Process the entire folder

You can simply use the bash for loop to process the all the images contains in a folder.

for file in original_images/*; do python AUTOCROP $i -o cropped_images; done

Since the parameter auto-detection doesn't work well on some of the plates, the below is the recommended processing step:

1. Since most of the plates are considered to be medium growing plate. We first run the for loop with the -m M option.
2. Manually check the images that are failed and delete them.
3. Run the for loop again with the -m L option. Delete the failed images.
4. Run the for loop again with the -m H option. Delete the failed images.
5. Manually crop image by GIMP

Try out on the notebook

A jupyter notebook test.ipynb was also uploaded to the repo. You can try to fine-tune the program by yourself. Please keep the lines importing the library to make it works. Especially the line which imports the autocrop functions.

import matplotlib.pyplot as plt
from types import SimpleNamespace

from autocrop.autocrop import *

Requirements

• Python >= 3.7
• opencv >= 4.8.0
• largestinteriorrectangle >= 0.2.0
• timeout-decorator >= 0.5.0

Use the environment.yml to install all the required packages:

conda env create -f environment.yml