ferrygun/pdftableextract

PDFTableExtract

PDFTableExtract

PDFTableExtract

Build a parser to extract the table in PDF document with RetinaNet https://medium.com/@djajafer/pdf-table-extraction-with-keras-retinanet-173a13371e89

pdftableextract's Issues

Missing .h5 file

I tried to follow the tutorial but i can neither clone the RN.git file from your repository nor find the resnet50_coco_best_v2.1.0.h5 file anywhere else.

Is there a follow-up to this issue? Or is the code uploaded somewhere else?

Impossible to use this!

Its impossible for others to follow what you have done in your "tutorial"!

Example:
model = models.load_model(model_path, backbone_name='resnet50')

• What package do you need to import in order to load this model?????

And in your "Now we are ready to run the prediction" there are several packages which need to be loaded, but it aint said anything about package names!

Also the "build_logos.py" is fail, because you need to remove .FILETYPE in the .txt files, and in your "tutorial" its pointed out that .FILETYPE has to be included.

Can you please make a complete tutorial which is possible to follow, step-by-step?

unable to load the trained model ,please help me on this concern

-------------------errrorr-------------------------------------
TypeError Traceback (most recent call last)
in ()
8 #print("path:", model_paths)[-1]
9 #from keras_retinanet import models
---> 10 model = models.load_model(model_paths)
11 model = models.convert_model(model)

2 frames
/usr/local/lib/python3.7/dist-packages/h5py/_hl/base.py in is_hdf5(fname)
39 """ Determine if a file is valid HDF5 (False if it doesn't exist). """
40 with phil:
---> 41 fname = os.path.abspath(fspath(fname))
42
43 if os.path.isfile(fname):

TypeError: expected str, bytes or os.PathLike object, not list

--------------code used ----------------------

from keras_retinanet import models
from glob import glob
#model_paths = glob('/content/Face-mask-detector-using-RetinaNet-model/keras-retinanet/keras-retinanet/snapshots/_pretrained_model.h5')
#model_paths = glob('snapshots/resnet50_csv_0*.h5')
model_path='/content/keras-retinanet/snapshots/_pretrained_model.h5'

#print("path:", model_paths)[-1]
#from keras_retinanet import models
model = models.load_model(model_paths, backbone_name='resnet50')
model = models.convert_model(model)

--------------tried on this issue -------------

tried to install h5py .but still issue is not solved .
pip install h5py==2.10.0

error in git clone https://github.com/ferrygun/RN2.git

Hi, it seems https://github.com/ferrygun/RN2.git does not exist anymore? I ran into the following errors:

Cloning into 'RN2'...
fatal: could not read Username for 'https://github.com': No such device or address

Could you help fix this? BTW, excellent work in PDFTableExtract!! Thank you so much!

issue in installing setup.py

Hello,

When I tried !pip install python setup.py build_ext — inplace , I am receiving the below error

Error
ERROR: Invalid requirement: '—'

Training Issue

Hello,
I have followed each instruction but while training it do not proceed till the end.
Please Help

Requesting Video

Hello ferrygun,

Good work.
could you make some video on this "https://djajafer.medium.com/pdf-table-extraction-with-keras-retinanet-173a13371e89", it will be really helpful for me to explore the same.

Thank you

I am getting 0 precision during retinanet-evaluate in TrainOCR

Hi,

I downloaded some receipts from internet and manually labeled and created annotations. Everything seems to be going as per plan, until I noticed 0 precision while evaluating retina-net output weight file:

I can see for you precision is about .83. What might be wrong?