Hello trend,
This is Satya Mallick from LearnOpenCV.com.
In today's post, we will learn about building a Custom Semantic Segmentation model for Scanning Documents using PyTorch. In our last post on Document Scanning, we had discussed a grabcut based segmentation approach which had some limitations. We are able to train a more robust Deep Learning based model called Deeplabv3 in PyTorch. We also show how to deploy it as a streamlit web application.
DeepLabv3 is a Semantic Segmentation model that classifies each pixel in the image to a particular class. Check out this video on Image Segmentation if you would like to know more about the different types of Segmentation and the differences between them.
In today's post, you will learn the following:
- Creating synthetic data to augment the dataset.
- Creating custom dataset classes in PyTorch.
- Fine-tuning DepLabv3 with custom loss functions.
- Deploying the application using streamlit.
Without further ado, let's get started.
| Custom Document Segmentation with DeepLabv3 |
You can check out the code by clicking on the button below, and star us on GitHub to say thanks!
| Download Code (GitHub) |
🎥 Introducing Blogpost Highlights Video
As suggested by many of you, we are trying out publishing an introductory video to give you a bird's eye view of the post. Please share your feedback to let me know if you'd like to see these short accompanying videos to introduce and summarize the post going forward. Hope you like it!
No comments:
Post a Comment
Note: Only a member of this blog may post a comment.