Image Document Classification using LayoutLM | Document understanding | смотреть онлайн

Documents in form of PDF or Images are available in the Financial domain, FMCG domain, healthcare domain, etc. and when documents are huge in numbers, it becomes challenging to classify them into a certain category as documents just do not vary with image patterns, it also varies with respect to textual context present in it. Hence, acknowledging both image and text information into a single model and classifying them into the right category makes it a more functional and appropriate approach to achieving a good classifier model instead of just fine-tuning simple image classifiers like VGG, ResNet, etc.

Video explains about LayoutLm Model fine-tuning to train on the custom dataset, using image and text information together to classify the document into the right category.

✅ LayoutLM Annotation Tool: https://youtu.be/r1aoFj974FU
✅ Signup for Annotation Tool: https://ubiai.tools/Signup?utm_source=youtube&utm_medium=video&utm_campaign=KS_campaign&utm_content=signup

1. Notebook: https://github.com/karndeepsingh/Extract_key_information_Document_understanding/blob/main/Fine_tuning_LayoutLMv2ForSequenceClassification_on_RVL_CDIP_(using_LayoutLMv2Processor).ipynb
2. Label Studio: https://labelstud.io/blog/Improve-OCR-quality-with-Tesseract-and-Label-Studio.html
3. LayoutLMv2 Paper: https://arxiv.org/abs/2012.14740v4
4. FUNSD Dataset: https://guillaumejaume.github.io/FUNSD/
5. https://github.com/NielsRogge/Transformers-Tutorials
6. LayoutLM Preprocessing file: https://github.com/UBIAI/layout_lm_tutorial
7. https://github.com/katanaml/sparrow


Connect with me on :
1. LinkedIn: https://www.linkedin.com/in/karndeepsingh/

2. Telegram Group: https://telegram.me/datascienceclubachievers

3. Github: https://www.github.com/karndeepsingh

#datascience #nlp #deeplearning #documentunderstanding