The advancement of computer vision to a new level has made it possible for laboratory robots to explore the physical environment. Even while this field makes strides, robots still have trouble figuring out their surroundings. Scene categorization is an important first step in gaining insight into the environment. A scene that can be classified at both the low and high levels. The primary function of automated surveillance applications such as indoor orientation, pedestrian identification, semantic categorizations, etc. is played by scene classification. DL approaches, and convolutional neural networks (CNN) in particular, have become a frequent option for scene categorization because they can remove features automatically and without the burden of doing so manually. CNNs with DL built in vastly outperform their predecessors. In order to create multi-layered learning models of neuron changes, it makes use of visual technology. In this work, a model for scene categorization that relies on characteristics extracted from deep CNN. The accuracy of scene classification has been improved by the application of transfer learning. To reduce parameter space and improve the feature quality, deep residual network is introduced as the feature extractor. From these kinds of non-technical encounters in the real world, it may gain insight into the reasons why transfer learning is possible. The goal of transfer learning is to apply the values of learned parameters from one model to another that has not yet been trained. This paper provides a brief introduction to Scene Classification, before moving on to IndoorOutdoor Classification and the different deep learning techniques employed therein

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