/Imagenet_annotator

Rapid Imagenet-style Annotation Generator only for Classification

Primary LanguagePython

Rapid Imagenet-style Annotation Generator (Classification)

This is an effective tool to prepare annotation sample with text annotation. If you want to filter using folder, check this repo.

  1. setup
conda create -n ann python=3.11
conda activate ann
git clone https://github.com/ccomkhj/Imagenet_annotator.git
cd Imagenet_annotator
pip install -r requirements.txt
  1. prepare config file config/class.yaml
noplants: 0
healthy: 1
unhealthy: 2
  1. run

if you want to create annotation only

python createClsAnns.py -i {image directory}

if you want to copy images

python createClsAnns.py -s -i {image directory}

if you want to copy images and save them based on the class name

python createClsAnns.py -sf -i {image directory}

if you want to start from specific index, (it is useful if you begin with the previous work.)

python createClsAnns.py -s -b {index to begin with} -i {image directory}

if you have a list of keywords that should be included.

python createClsAnns.py -s -k {keyword} -i {image directory}

in config/key.yaml

abc:
 - G8T1-K001-2173-0FPQ
 - G8T1-K001-2173-0FUF
def:
 - xxx
 - yyy

  1. type relevant class number per image

  2. if you made mistake, type 'x'. It is logged that you made mistake in which image file.

  3. if you want to quit, type 'q'.

Output

result/ann_{epoch time}.txt

result/log_{epoch time}.txt

Wanna split train and validation set?

w/o spliting folder

python split.py -i {image directory} -a {annotation text file} -r 0.8

w/ spliting folder

python split_folder.py -i {image directory} -r 0.8