- Clone forked repo to local machine
- Create virtual environment and active it
conda create -n <venv-name> python=3.6
conda activate <venv-name>
- Install dependencies, by run the followings commands
pip install opencv-python
conda config --add channeds conda-forge
conda install imgaug
- Change directory to dataset directory
cd <project-root-dir>/dataset/<dataset-name-dir>
- Place dataset into that directory
- Change directory to
<project-root-dir>
- Run the following command
python bucketing -dn <dataset-name>
- The result will be saved at
<project-root-dir>/buckets/<dataset-name>/<datasetname-bucket_X>
- Open http://dataturks.iapp.co.th/projects/login
- Log in with
Email : iwachirawit@iapp.co.th
- Create new dataset
- Upload compressed bucket e.g. upload
<project-root-dir>/buckets/<dataset-name>/<datasetname-bucket_X>.zip
- Open http://dataturks.iapp.co.th/projects/login
- Log in with
Email : iwachirawit@iapp.co.th
- Select desired dataset
- Click on
options
button at upper right corner, then click download you
After downloading finished you will get json file as a result
- Place downloaded json file under
<project-root-dir>/dataturks/json/<datasetname>
After that, you need to convert tagged data from json format into PascalVOC format
- Change directory to
<project-root-dir>
, run the following commands
dataturks_to_PascalVOC.py -dn <datasetname>
- You will get dataset in PascalVOC format at /dataturks/pascalVOC>/
- Change project root directory
cd <project-root-dir>
- Run
python crop_lp.py -dn <datasetname> —height <desired height>
- You will get the result under
<project-root-dir>/cropped_lps/<datasetname>