https://towardsdatascience.com/borderless-tables-detection-with-deep-learning-and-opencv-ebf568580fe2

Steps to install with virtualenvwrapper:

1. Find the location of Python with a specific version:

$ which python
/home/USERNAME/.virtualenvs/tf/bin/python

The path to Python is a softlink. Use the following command to find where the original file is:

$ namei /home/USERNAME/.virtualenvs/tf/bin/python
f: /home/USERNAME/.virtualenvs/tf/bin/python
 d /
 d home
 d USERNAME
 d .virtualenvs
 d tf
 d bin
 l python -> /usr/bin/python3.9
   d /
   d usr
   d bin
   - python3.9

Use the path you found /usr/bin/python3.9 to create a new virtual environment:

mkvirtualenv -p /usr/bin/python3.9

2. Install required modules

pip install tensorflow tf_slim cython

3. Clone Tensorflow models

git clone https://github.com/tensorflow/models.git

4. Install pycocotools

sudo apt install python3.9-dev
pip install git+https://github.com/philferriere/cocoapi.git#subdirectory=PythonAPI

5. Install CUDA

Follow the tutorial from here.

5. Prepare project dirs

cd models/research
cp object_detection/packages/tf2/setup.py .
python setup.py install
python object_detection/builders/model_builder_tf2_test.py