Weather Classification with Transfer Learning - InceptionV3, MobileNetV2 and ResNet50
Patryk Młodzianowski
Abstract
Abstract Weather recognition is a common problem for many branches of industry. For example self-driving cars need to precisely evaluate weather in order to adjust their driving style. Modern agriculture is also based on the analysis of current meteorological conditions. One of the solutions may be a system detecting weather from image. Because any special sensors are needed, the system should be really cheap. Thanks to transfer learning it is possible to create image classification solutions using a small dataset. In this paper three weather recognition models are proposed. These models are based on InceptionV3, MobileNetV2 and ResNet50 architectures. Their efficiency is compared and described.
Topics & Concepts
Transfer of learningComputer scienceImage (mathematics)Artificial intelligenceContextual image classificationMachine learningSmart Agriculture and AIFire Detection and Safety SystemsWater Quality Monitoring Technologies