Image De-Raining Enhancement Tool

In this fast-paced technological day and age, computer vision systems significantly and positively impact our daily lives. They are the basis of many systems that are now crucial in our everyday lives like healthcare and medical imaging, object recognition and visual search, autonomous vehicles, industrial automation, and robotics, as well as accessibility and assistive technologies. These are just a few examples of how computer vision has become indispensable. These systems primarily depend on the quality of images, where the slightest distortion could lead to catastrophic malfunctions. Therefore, in this research, a developed Spinning Detail Perceptual Generative Adversarial Network (SDP-GAN) was proposed for single image de-raining, aiming to enhance the performance of computer vision-based systems. The presented algorithm is a Generative Adversarial Network (GAN) based network and it is built using two subnetworks: the rain streak generative network G and the discriminative network D. To further enhance the perceived quality of the derained images, modifications have been added to find the most appropriate values for the model's hyperparameters to allow it to generate the most optimal results. The batch size, number of epochs, optimizers, filters, and loss functions have all been altered to ultimately improve the performance of the model; consequently, improving the overall de-raining process. Numerous tests on commonly used synthetic datasets validate that the proposed model significantly outperforms several state-of-the-art approaches.