Document Layout Analysis App

Simple docker deployment of document layout analysis using detectron2. Trained on Publaynet dataset.

Backend is built in Flask

Frontend is built in React

Steps to build

• Run docker-compose up --build from the main folder
• Mask RCNN R101 will be auto downloaded
• Type localhost:3000 in browser and upload and image

(Optional) To use model other than Mask RCNN R101

• Download pre-trained model refered in https://github.com/hpanwar08/detectron2 (default Resnet101 model can be downloaded from MaskRCNN Resnet101 FPN 3X )
• Copy the downloaded model in backend folder
• Download config yaml refered in https://github.com/hpanwar08/detectron2 (default Resnet101 config yaml is already present in backend folder)
• Copy the downloaded config yaml in backend folder
• Update the model_config.yaml file with model file name and config yaml file name (not required if using default values)
• Run docker-compose up --build from the main folder
• Type localhost:3000 in browser and upload and image

Note: It takes time to build the image