Develop a shared weights model to detect table and text inside the table at the same time

[sgrpanchal31]

We are using two different model, one for detecting tables and the other to detect text inside the table. Though their outputs are different, calculations in the initial layers of the models are the same but still, they are being done twice. Therefore, developing such a model will reduce the inference time by at least 50%.

[mariemabc]

did you try sharing weights between RPN and ctpn ?