Development of deep neural network for document layout analysis over publaynet dataset

Development of deep neural network for document layout analysis over publaynet dataset.

You have to train on the publaynet dataset and develop a model for detecting the text, title, image,

He/she has to use CNN or RNN or a combination of both or transformers to extract features from document images.

He has to use transfer learning to conduct fine tuning of model pretrained on Publaynet Dataset.

He has to compare different deep learning architectures like CNN, faster RCN, RNN, and transformer based models. He can use facebook detectron2 or Layoutparser for fine tuning

The model will be trained on Publaynet dataset and its performance will be evaluated using metrics such as accuracy, F1 score, recall and precision.

The performance of developed network can also be compared to existing annotation methods using statistical tests such as T tests and annova.

Proper graphs about the iterations has to be plotted.

[login to view URL]

Python Datatiede Machine Learning (ML) Pilvitietotekniikka

Projektin tunnus: #36654092

Tietoa projektista