Página inicial Explorar Ajuda
Entrar
ICILearn
/
learnopencv
mirror de https://github.com/spmallick/learnopencv.git
Observar 2
Favorito 0
Fork 0
Arquivos Issues 0 Wiki
Branch: master
Branches Tags
learnopencv
/
Document-Scanner-Custom-Semantic-Segmentation-using-PyTorch-DeepLabV3
Yigit 989a8dd515 Update Document_extraction.ipynb 2 meses atrás
..
app_images d296c91ec9 assets added for Custom document segmentation 2 anos atrás
test_images d296c91ec9 assets added for Custom document segmentation 2 anos atrás
Custom_training_deeplabv3_mbv3.ipynb d296c91ec9 assets added for Custom document segmentation 2 anos atrás
Document_extraction.ipynb 989a8dd515 Update Document_extraction.ipynb 2 meses atrás
PROCEDURE.md d296c91ec9 assets added for Custom document segmentation 2 anos atrás
README.md 57b5cefde5 Updated all `.md` files to contain newest image 2 anos atrás
app.py d296c91ec9 assets added for Custom document segmentation 2 anos atrás
create_dataset.py d296c91ec9 assets added for Custom document segmentation 2 anos atrás
download_google_images.py d296c91ec9 assets added for Custom document segmentation 2 anos atrás
generate_doc_set.py d296c91ec9 assets added for Custom document segmentation 2 anos atrás
resizer.py d296c91ec9 assets added for Custom document segmentation 2 anos atrás
split_dataset.py d296c91ec9 assets added for Custom document segmentation 2 anos atrás

README.md

Document Scanner: Semantic Segmentation using PyTorch-DeepLabV3

This repository contains code for the blog post Document Scanner: Semantic Segmentation using PyTorch-DeepLabV3.

In this post, you will learn the following,

  • Generate Synthetic Dataset.
  • Create a Combo loss function training a Custom Image Segmentation Model.
  • Train a Document Segmentation model using DeepLabv3-PyTorch.

Dataset and Trained Model Download Links

  1. Resized final dataset
  2. Model - MobileNetV3-Large backend
  3. Model - Resnet50 backend

download

Document Scanner Application

AI Courses by OpenCV

Want to become an expert in AI? AI Courses by OpenCV is a great place to start.