PG-HWLD v1.0

You can download our dataset from here.

PG-HWLD v0.1

Development, non-final version of the dataset.

Tesseract binary

Unpack contents of this zip and place it in models/pytesseract/bin

v1.0.0

dataset-EMNIST has to be unpacked in resources/datasets
dataset-multi-person and dataset-single-person have to be unpacked in resources/datasets/unpacked

You can then create further datasets by changing the margin value and running "ImageCropper", which creates square images centered on the character with preset margins on all 4 sides. (eg. margins 5, 10, 15)

After creating all datasets run "ImageTransformer" to convert all images in "resources/datasets/unpacked" to 28x28 denoised normalized model inputs. This may take up to 30 mins (no MT implemented).

TODO: Further simple data augmentation, Data Shuffler for testing sets

Dataset multi-person cropped-10 augmented

partial-dataset

Merge branch 'imagetransformer-mp'

v0.0.0

Those are zip files I found in our current project. You should put them under resources/datasets/archives.
And obviously, unpack them into resources/datasets/unpacked.
Good luck

Full Changelog: https://github.com/blazej222/AI-Project/commits/experimental