conda create -n tfobjdetection pip python=3.9
conda activate tfobjdetection
pip install --ignore-installed --upgrade tensorflow==2.11.0
- Download Dataset via kaggle API or manually from KAGGLE
- Follow the instructions for Downloading via Kaggle API
- Run
kaggle datasets download -d andrewmvd/car-plate-detection
in the terminal where the project folder is
- Run
python ./1.PrepareDataset.py
locally
- Upload file
ANPR_and_EasyOCR_ColabRun_v1.ipynb
in Google Colaboratory - Remember to Upload
archive.tar
file of prepared dataset in the step 2 when runningANPR_and_EasyOCR_ColabRun_v1.ipynb
in the directoryTensorflow/workspace/images
- Note down the latest checkpoint in the folder
Tensorflow\workspace\models\CUSTOM_MODEL_NAME\
e.g.ckpt-100
. This will be required to enter in scripts3.DetectFromImage_EasyOCR.py
,4.DetectFromRealTimeFeed_EasyOCR.py
,5.DetectFromVideos_EasyOCR.py
,app.py
whereLOAD_CHECKPOINT = 'ckpt-101'
- Afer training, Download the compressed file of trained model in the project folder and uncompress it.
- Note:
3.DetectFromImage_EasyOCR.py
,4.DetectFromRealTimeFeed_EasyOCR.py
,5.DetectFromVideos_EasyOCR.py
these are optional files.
- Run
streamlit run app.py
- Go Canva