Currently Updating the repository

SurgicalSceneRestoration

Dataset for "One step surgical scene restoration for robot assisted minimally invasive surgery" method applied on Knee Arthroscopy. The model learns image restoration methods namely denoise, deblur,and color corrections. For other minimally invasive surgical (MIS) procedures or endoscopic images e.g. wireless capsule endoscopy, The model can be pre-trained on the following published dataset and can be used
fine-tuned on small endoscopic images.

Directory Structure

dataset/image: Real world endoscopic surgical images. [Updating]

dataset/synthetic_gt: Synthetic rendered surgical images.

dataset/clean: Clean restored denoised deblurred and color corrected surgical images.

   Real world image                             Restored images

image resize function

   import os
   import sys
   import numpy as np
   import cv2
   import csv
   import os


   rd_dir='./PATH_TO_READ_DIRECTORY/'
   save_dir='./PATH_TO_SAVE_DIRECTORY/'

   def make_dir_(dir_name):
       if not os.path.exists(dir_name):
           os.makedirs(dir_name)
           return True
       else:
           return False

   def resize_images():
       file_list= os.listdir(rd_dir)
       for c in range(len(file_list)):
           name=file_list[c]
           I = cv2.imread(rd_dir+name) # read the image
           # resize either 256x256 or 384x384
           #image = cv2.resize(I , (256, 256))
           image = cv2.resize(I , (384, 384))
           cv2.imwrite(save_dir+name, image)

Citation:

   Ali, Shahnewaz, Yaqub Jonmohamadi, Davide Fontanarosa, Ross Crawford, and Ajay K. Pandey. "One step surgical scene restoration for robot assisted minimally invasive surgery." Scientific Reports 13, no. 1 (2023): 3127.

Doi

   https://doi.org/10.1038/s41598-022-26647-4