Scenic buildings Recognition Based on MobileNetV1 Network and Transfer Learning

The recognition of images in natural scenes based on mobile devices has become mainstream. The cost of recognizing images through mobile phone photography is low and the operation is simple. A method based on MobileNetV1 and transfer learning is proposed to identify scenic buildings in China due to the small sample size and low recognition accuracy. Firstly, collect a dataset of buildings from famous scenic spots in China, and use methods such as flipping, flipping, scaling, brightness adjustment, adding noise, and pixel translation to improve the diversity of the data samples and avoid overfitting the model. Use MobileNetV1 to pre train the model on the ImageNet dataset; Then, based on the pre trained results, a MobileNetV1 network for scenic spot building recognition was constructed. The weights of the pre trained MobileNetV1 network model were transferred to the constructed MobileNetV1 network for model weight initialization, thereby reducing training costs. The outcomes of the experiment demonstrate that using transfer learning enables the model to complete training quickly, and the accuracy is greatly improved compared to non-transfer learning methods. Research has shown that using methods based on MobileNetV1 and transfer learning effectively solves the problems of small sample size and low accuracy in scenic building recognition, providing a new method for scenic building recognition.