Troubleshooting tensorflow model project

Problems running the object detection tutorial

run object detection jupyter notebook

cd object_detection
set PYTHONPATH=D:\ming\git\tensorflow-models;D:\ming\git\tensorflow-models\research;D:\ming\git\tensorflow-models\research\slim
jupyter notebook object_detection_tutorial.ipynb

problem

AttributeError: module 'tensorflow._api.v1.compat' has no attribute 'v1'

troubleshoot
conda list shows tensorflow 1.15
jupyter notebook print(tf.version) shows 1.12 conda does not start jupyter with right env
solution
reinstall anaconda
https://repo.anaconda.com/archive/Anaconda3-2019.10-Windows-x86_64.exe

conda create -n tensorflow1 pip python=3.5
conda activate tensorflow1
conda install -c anaconda protobuf
python -m pip install --upgrade pip
pip install tensorflow-gpu==1.15
pip install --user Cython
pip install --user contextlib2
pip install --user pillow
pip install --user lxml
pip install --user jupyter
pip install --user matplotlib

---

problem
jupyter notebook kernel error

File "C:\Users\ming\AppData\Roaming\Python\Python37\site-packages\jupyter_core\paths.py", line 359, in win32_restrict_file_to_user
    import win32api
ImportError: DLL load failed: 找不到指定的程序。

solution //make sure to use python 3.5
pip uninstall tensorflow-gpu
//conda install tensorflow-gpu==1.14
//search on http://anaconda.org
//Package certifi conflicts for:
//python=3.5 -> pip -> requests -> certifi[version=’>=2016.09|>=2017.4.17’]
//tensorflow-gpu==1.14 -> tensorflow==1.14.0 -> python=3.6 -> pip -> requests -> certifi[version=’>=2016.09|>=2016.9.26|>=2017.4.17’]
check C:\Users\ming.conda\environments.txt
conda create -n tensorflow1 pip python=3.5
conda activate tensorflow1
conda install -c anaconda protobuf
//conda install tensorflow-gpu
//python -m pip install –upgrade pip
pip install tensorflow-gpu==1.15
pip install –user Cython
pip install –user contextlib2
pip install –user pillow
pip install –user lxml
pip install –user jupyter
pip install –user matplotlib

set PYTHONPATH=D:\ming\git\tensorflow-models;D:\ming\git\tensorflow-models\research;D:\ming\git\tensorflow-models\research\slim
set PYTHONPATH=D:\ming\git\models;D:\ming\git\models\research;D:\ming\git\models\research\slim
jupyter notebook object_detection_tutorial.ipynb

---

problem with anaconda

 File "C:\Users\ming\AppData\Roaming\Python\Python37\site-packages\jupyter_core\paths.py", line 359, in win32_restrict_file_to_user
    import win32api
ImportError: DLL load failed: 找不到指定的程序。

Add \Anaconda3 and \Anaconda3\Script to Path in Environment variables. add \Anaconda3\Library\bin

(tensorflow1) C:\Users\ming>jupyter kernelspec list
Available kernels:
python3 D:\Anaconda3\envs\tensorflow1\share\jupyter\kernels\python3

(tensorflow1) C:\Users\ming>python -m ipykernel install –user
Installed kernelspec python3 in C:\Users\ming\AppData\Roaming\jupyter\kernels\python3

new install C:\Users\ming\Anaconda3
delete
C:\Users\ming\AppData\Roaming\jupyter
C:\Users\ming\AppData\Roaming\Python
C:\Users\ming
.anaconda
.conda
.condarc
.ipynb_checkpoints
.ipython
start base terminal
(base) C:\Users\ming>jupyter kernelspec list
Available kernels:
python3 C:\Users\ming\Anaconda3\share\jupyter\kernels\python3

anaconda navigator create env tensorflow1
install notebook
(tensorflow1) C:\Users\ming>jupyter kernelspec list
Available kernels:
python3 C:\Users\ming\Anaconda3\envs\tensorflow1\share\jupyter\kernels\python3

---

problem
kernel crashes on load label
label_map_util.load_labelmap(PATH_TO_LABELS) solution
check that object_detection package has been installed

cd research
pip install . 

object_detection jupyter notebook uses tf2 so create tf2 environment
“!pip install -U –pre tensorflow=="2.*"”

conda create -n tensorflow2 pip python=3.5 conda activate tensorflow2
conda install -c anaconda protobuf
python -m pip install –upgrade pip

pip install tensorflow-gpu==2.0 problems when running show_inference(detection_model, image_path)

UnknownError: 2 root error(s) found.
  (0) Unknown:  Failed to get convolution algorithm. This is probably because cuDNN failed to initialize, so try looking to see if a warning log message was printed above.
	 [[node FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_0/BatchNorm/batchnorm/mul_1 (defined at C:\Users\ming\Anaconda3\envs\tensorflow2\lib\site-packages\tensorflow_core\python\framework\ops.py:1751) ]]

Then, once inside the environment, install TensorFlow using CONDA rather than PIP: conda install tensorflow-gpu use navigator to install will download cudnn

pip install –user Cython
pip install –user contextlib2
pip install –user pillow
pip install –user lxml
pip install –user matplotlib

pip install –user jupyter
use navigator to install

cd research
pip install .

set PYTHONPATH=D:\ming\git\models;D:\ming\git\models\research;D:\ming\git\models\research\slim
python -m ipykernel install –user
Installed kernelspec python3 in C:\Users\ming\AppData\Roaming\jupyter\kernels\python3
jupyter kernelspec list
Available kernels:
python3 C:\Users\ming\AppData\Roaming\jupyter\kernels\python3
kernel.json
“C:\Users\ming\Anaconda3\envs\tensorflow2\python.exe”,
this is okay it matches env python
jupyter notebook object_detection_tutorial.ipynb

---

conda install -c conda-forge pycocotools
https://github.com/conda-forge/pycocotools-feedstock
conda config –add channels conda-forge
Once the conda-forge channel has been enabled, pycocotools can be installed with:
conda install pycocotools

---

kept getting

UnknownError: 2 root error(s) found.
  (0) Unknown:  Failed to get convolution algorithm. This is probably because cuDNN failed to initialize, so try looking to see if a warning log message was printed above.
	 [[node FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_0/BatchNorm/batchnorm/mul_1 (defined at C:\Users\ming\Anaconda3\envs\tensorflow2\lib\site-packages\tensorflow_core\python\framework\ops.py:1751) ]]

check out https://github.com/tensorflow/models/tree/r1.13.0
conda activate tensorflow1
change to tf 1.13
conda install tensorflow-gpu==1.13
make sure
python object_detection/builders/model_builder_test.py
worked then
jupyter notebook object_detection\object_detection_tutorial.ipynb

if images do not appear edit D:\ming\git\models\research\object_detection\utils\visualization_utils.py

#import matplotlib; matplotlib.use('Agg')  # pylint: disable=multiple-statements
#import matplotlib.pyplot as plt  # pylint: disable=g-import-not-at-top