machine-learningimage-processingdeep-learningcomputer-visionocr
Text Image Binary Classifier without deep learning or tesseract
I have 20k small label images and each image has the word "Back" or "Front".
Image resolution is all (200px, 25px)
I can classify these images 100% accuracy with tesseract_OCR.
txt = pytesseract.image_to_string(img, lang='eng')
if "Front" in txt:
return "Front"
if "Back" in txt:
return "Back"
problem is, it is too slow(1 hour for 20k images) and need to install OCR packages.
I know even 3 layer Simple CNN also works well for it, but I think this problem seems to be solvable with simple algorithm without difficult techniques.
Can you recommend a new approach for me?
thank you.
Solution
I used the template_match function of OpenCV to compare whether the "Front" image saved as a sample matches the "Back" image.
I succeeded in reducing the time to about 1 minute.
https://docs.opencv.org/4.x/d4/dc6/tutorial_py_template_matching.html
- Am I implementing my perceptron with backpropagation correctly?
- Issue setting up SciKeras model
- Should the data in batch be balanced?
- Multi Step Prediction Neural Networks
- Train and test splits by unique dates, not observations
- Human segmentation fails with Pytorch, not with Tensorflow Keras
- Keras multioutput custom loss with intermediate layers output
- Isolation Forest Sklearn for 1D array or list and how to tune hyper parameters
- Query padding mask and key padding mask in Transformer encoder
- Masking and computing loss for a padded batch sent through an RNN with a linear output layer in pytorch
- Why does nn.Linear(in_features, out_features) use a weight matrix of shape (out_features, in_features) in PyTorch?
- Facial Expression Recognition Data Preparation for CNN
- Machine Learning: Image classification into 3 classes (Dog or Cat or Neither) using Convolutional NN
- PyTorch "Caught IndexError in DataLoader worker process 0", "IndexError: too many indices for array"
- What does "unsqueeze" do in Pytorch?
- Importing a Model with Scikit Learn on Vertex
- Should nested modules with shared weights be an nn.Module object parameter or not?
- Convert array of indices to one-hot encoded array in NumPy
- All intermediate steps should be transformers and implement fit and transform
- How do I crop the solar panels captured by drone?
- Performance issue with gradient-bandit agent
- How to concatenate a tensor to a keras layer along batch (without specifying batch size)?
- How do I import KerasRegressor in Google Colab?
- Text Image Binary Classifier without deep learning or tesseract
- Paralellizing a natively single-batch pytorch model
- What is the problem with my implementation of the cross-entropy function?
- `CrossEntropyLoss()` in PyTorch
- How to use Machine Learning to find the pattern customer profile?
- Why is ReLU a non-linear activation function?
- Extracting an information from web page by machine learning