2022 Developer Survey is open! Take survey.

Hot answers tagged

8 votes

Fake News Detection problem

This is a very ambitious project. First it's important to realize that ML cannot really solve this kind of problem in general, it can only help detect the posts which are likely fake news (see for ...
user avatar
  • 20.9k
7 votes

How can you build a model that extracts data out from receipts?

The simplest pipeline would be to do the following: OCR Named Entity Extraction Entity Disambiguation OCR This is basically transforming your receipts into plain text. If you have scans (pictures) ...
user avatar
5 votes
Accepted

Document Layout Analysis - state-of-the-art?

After some research, I came across ICDAR (International Conference on Document Analysis and Recognition), which is taking place biannually and seems to be the most complete and up-to-date source for ...
user avatar
  • 211
4 votes

How to improve OCR (Scanning) results?"

Build a dictionary of common words that frequently appear in these documents (e.g., MEDICAL, SEX, AGE, etc). Then, for each word in the output from OCR, check whether it is similar to a word in your ...
user avatar
  • 2,968
3 votes

Evaluating information extraction from structured documents

I have worked with Structured text using OCR. OCR is prone to errors even while reading the content and a slight change in string arrangement would lead to false positives. I used Cosine Similarity ...
user avatar
  • 339
3 votes

How to pass features extracted using CNN into RNN?

1 and 2. You are in the right direction, you need to extract the features using a CNN, then instead of predicting the class you want to reshape the last layer of features and feed it directly into the ...
user avatar
  • 7,428
2 votes
Accepted

how can I solve label shape problem in tensorflow when using one-hot encoding?

Since your data are already in a one-hot encoding, you can use tf.nn.softmax_cross_entropy_with_logits(), which expects an input of shape ...
user avatar
  • 136
2 votes
Accepted

what is the loss function in char recognition using Tensorflow?

The loss function is correct, you just need to convert categorical variables into numerical representations using one-hot vector encoding. Please take a look at this.
user avatar
  • 64
2 votes
Accepted

Can I train two stacked models end-to-end on different resolutions?

I would suggest you to split your problem into two: How to train? How to make inference when resources are limited? Very common pattern is to train model on large data (by renting AWS server, for ...
user avatar
2 votes

Extract text from a image - OCR

I have used tesseract for similar tasks. I can give you few recommendation. You can choose the best that works for you. Extract the parameter values by finding the exact locations they appear. If ...
user avatar
  • 171
2 votes
Accepted

Pretrained handwritten OCR model

Discover open-source deep learning code and pretrained models at Model Zoo These are pre-trained sources available in the Github. Handwritten Text Recognition with TensorFlow More Handwriting OCR ...
user avatar
  • 2,301
2 votes

optical chemical structure recognition from images

What way I could follow to achieve this? Create a dataset with image-label pairs. Create a classification model. I don't think ...
user avatar
2 votes

how can i detect medicine name and info(use and contents) by using medicine wrappers

For data, you can search on IEEE Dataport , Kaggle. For detecting medicine names and other info trained deep learning models like CNN or you can also perform fine-tuning from the existing model.
user avatar
  • 83
1 vote

How to detect medicine name from the medicine wrapper

The Python Library Python-tesseract is an optical character recognition (OCR) tool for python. That is, it will recognize and “read” the text embedded in images. Python-tesseract is a wrapper for ...
user avatar
1 vote

License Plate OCR

Intel OpenVINO is used to optimise inference speed on edge devices. One of their demos was license plate detection. The link can be found here: https://docs.openvinotoolkit.org/2019_R1/...
user avatar
1 vote

How to create a model to recognize matching label and ROI with OCR

There are multiple ways to do it. Here is one technique: Run OCR on your image and search for your desired text . in this case it is Total. Get the pixels of that word - run ...
user avatar
  • 106
1 vote
Accepted

training neural nets for OCR

CTC can also work with totally random text (and thus without any "word pattern"). I know that because I trained an OCR network with CTC loss on random text which was then able to read non random text (...
user avatar
1 vote

Extracting text from few areas on product label

If the image will always be the same and your ROI will always be in the same location, then an RNN with CV2 is not necessary, you can mark the areas that you need to extract the texts. I even ...
user avatar
1 vote
Accepted

Machine learning algorithm which gives multiple outputs from single input

Keras functional API's are a way your can solve you problem. Using keras functional API, we can build models that resembles more like graphs such as this: In order to build a model like this, you ...
user avatar
  • 2,235
1 vote

How can I split an image into rectangles?

AWS Rekognition works very well for detecting typewritten text but not so well for detecting hand written symbols. You can see this is true in your example. What I would do is detect all the boxes ...
user avatar
  • 8,478
1 vote

How can you build a model that reads out receipts and invoices?

You should start by trying OCR. I am not sure why you are rejecting it without trying it. Start with Tesseract (free but not very good) and then try a commercial OCR as well (Abbyy is well regarded, ...
user avatar
  • 2,968
1 vote

High, constant training loss with CNN

I suspect you are facing a dying ReLU problem. Check the gradients for each layer and see if they are starting to become 0. Due to the way backpropagation works and a simple application of the chain ...
user avatar
  • 2,187
1 vote

What model to use for matching two datasets

Let the credit card statement be the ground truth, and the receipts be the noisy inputs. For a given line item, find the receipt with the smallest distance. If the distance is small enough, declare a ...
user avatar
  • 10.4k
1 vote

What is representation in optical character recognition?

The representation step is before the feature extraction step for exactly the reasons that they state. If you take the full image representation and go right to feature extraction, you have much ...
user avatar
  • 106
1 vote

What is the best approach for specified optical character recognition?

Optical character recognition is a well-studied problem with many possible solutions (ressources). CNNs have proven to work extremely well even for hand-written character recognition. Take a look at ...
user avatar
  • 571

Only top scored, non community-wiki answers of a minimum length are eligible