This program intends to extract the sift features from extremal regions. The stepwise manner mentioned as follow will guide you about the overall program execution.
Step 1: Try to execute the SIFT-points-detection-in-MSER.py file in terminal, before execution be sure to have opencv, matplotlib and all necessary packages that help in smooth execution of the program.
Step 2: Then install RNNLIB library from https://github.com/szcom/rnnlib, this library will help you to train the extracted SIFT features appeared in conjoined area of binary image and image mask.
Step 3: Pass (x,y) coordinates of selected SIFT features along with labels to RNN.
Saad Bin Ahmed, Saeeda naz, Muhammad Imran Razzak, Rubiyah Yusof, "A novel Dataset for English-Arabic Scene Text Recognition (EASTR)-42k and Its Evaluation using Invariant Feature Extraction on Detected Extremal Regions" in IEEE Access (2019).