Figure 2 | Scientific Reports

Figure 2

From: Preserving privacy in surgical video analysis using a deep learning classifier to identify out-of-body scenes in endoscopic videos

Figure 2

Model architecture of OoBNet. The input image is resized to 64 × 64 pixels and augmented with random rotation and contrast. Then it is fed to the deep neural network with a consecutive long short-term memory (LSTM) which outputs a probability like value whether the image is out-of-body or not. This probability is rounded at a 0.5 threshold to either 0 (inside the body) or 1 (out-of-body OOB).

Back to article page