Make --text-line-images debug option apply recognition preprocessing
#30
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Add `OcrEngine::prepare_recognition_input` method that returns an image with the same preprocessing applied as `OcrEngine::recognize_text` does before it feeds input into a model. This is useful for debugging scenarios where ocrs produces different / worse output than the PyTorch model training/evaluation tools.
`anyhow::Error` provides a better dynamic error type than `dyn Error` as it can capture context and be sent between threads. Using it here also enables propagating these errors in ocrs-cli which is already using anyhow.
Make the `--text-line-images` option save images with the same preprocessing applied as when preparing images to feed into the recognition model. This makes accuracy errors arising from preprocessing issues easier to debug. This preprocessing includes: - Resizing the image to 64px high and a max width of 800px - Extracting only the polygon containing the line's words, with other pixels masked off - Converting the image to grayscale
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Make the
--text-line-imagesdebug option apply the same preprocessing that is applied before lines are fed into the text recognition model. This includes:This makes this option more useful for debugging recognition accuracy issues, as problems arising from the preprocessing become visible.
In the process of doing this functions in ocrs which return dynamic errors were changed to use
anyhow::Errorrather thanBox<dyn Error>as the error type. This is more convenient to work with in ocrs-cli, which already used anyhow.