Denoised Autoencoder using DCNN Transfer Learning Approach
Richa Singh, Ashwani Kumar Dubey, Rajiv Kapoor · 2022 International Mobile and Embedded Technology Conference (MECON) · 2022
Many image denoising models have been designed for enhancing visibility of noisy images. In computer vision denoising of images is an important problem. A lot of emphasis is given to Machine Learning and Deep Learning approach in this regard. This work proposes the study of Deep Convolutional Neural network-based model VGG16 with the custom dataset of bad weather outdoor images via Transfer Learning. The sequential model 1 and model 2 is evaluated to have smooth image. The model architecture using Deep CNN is designed, and training of model is obtained using transfer learning. A model using autoencoder is designed. To decrease training time and perform better rectified Linear Unit (RELU) is used. Number of epochs identified to have increased performance in CNN. Further various optimizers are compared to have better accuracy. The estimation of performance such as RMSE and PSNR values are evaluated. The model is applied to single as well as 25 own outdoor images.