-
A classifier was trained to crop out handwritten/printed comments in the CLaMM 2016 [1] and 2017 [2] datasets.
-
CNN Model is adapted from [3].
-
Trained on 28 images from the 2016 dataset.
Install all Python requirements:
pip install -r requirements.txt
Run comment remover:
python annotation_remover.py --img_dir [clamm_img_dir] --plot [yes/no]
--img_dir specifies were the clamm images are and --plot if we should run an example on the first image found while making various debug plots.
To train the model see the train.py file and its main function for examples. However, model weights are provided. Model remover_model_v1_pad.keras was trained with padding while remover_model_v1.keras was not.
-
annotation_remover.py: includes all classes to run the pipeline
-
extract.py: extract contours from 28 cropped training images
-
train.py: train and test the CNN
-
test.py: tests the pipeline on 50 random CLaMM images
-
ocr.py: uses TrOCR to transcribe a comment
-
test_imgs_final.csv and test_imgs_final_2.csv: experiment results from 50 random images
-
history.json, scores.npy and model_summary.txt: epoch results and benchmarking from CNN training
[1] Cloppet, F., Eglin, V., Stutzmann, D., & Vincent, N. (2016, October). ICFHR2016 competition on the classification of medieval handwritings in latin script. In 2016 15th International Conference on Frontiers in Handwriting Recognition (ICFHR) (pp. 590-595). IEEE.
[2] Cloppet, F., Eglin, V., Helias-Baron, M., Kieu, C., Vincent, N., & Stutzmann, D. (2017, November). Icdar2017 competition on the classification of medieval handwritings in latin script. In 2017 14th IAPR International Conference on Document Analysis and Recognition (ICDAR) (Vol. 1, pp. 1371-1376). IEEE.
[3] Ahamed, P., Kundu, S., Khan, T., Bhateja, V., Sarkar, R., & Mollah, A. F. (2020). Handwritten Arabic numerals recognition using convolutional neural network. Journal of Ambient Intelligence and Humanized Computing, 11, 5445-5457.