A Survey on Indoor-Outdoor Scene Classification with Deep Learning Techniques

Deepika Bhardwaj, Vinod Todwal

Scene Classification, Indoor-Outdoor Classification, Transfer Learning, Deep Learning.

In the robotics and computer vision application, Scene classification is a fundamental challenge. A clear view or event with low& high-level featuresin Scene classification. Scene classification plays themain role in automated applications of surveillance comprising indoor orientation, pedestrian identification, semantic categorizations, etc. DL methods, especially CNN, which remove the features automatically and have no overhead for manual removal, have become a common solution for scene classification. DL, which is embedded in CNNs, improves its predecessors significantly. It uses graphic technologies for developing multilayered learning models of neuron transformations. A scene classification model was introduced dependsupon deep CNN features. The transfer learning method has been used to increase scene classification precision. Transfer learning is used to transfer control knowledge from a related domain to better the learner from one domain. We may learn about why transfer learning is feasible through real-world non-technical interactions. Transfer learning is aimed at using formerly trained models to transfer learning parameter values to new models. In this work, we outline Scene Classification its Pro & Cons of Diverse Scene Classification and after introduced Indoor-Outdoor Classification and various deep learning algorithms used in scene classification.