Detection of retinal diseases using an accelerated reused convolutional network

Convolutional neural networks are constantly being developed, some efforts improve accuracy, some increase speed, and some increase accessibility. Improving accessibility allows the use of neural networks in a wider range of tasks, including the detection of eye diseases. Early diagnosis of eye diseases and visiting an ophthalmologist ca n prevent many vision disorders. Because of the importance of this issue, various data sets have been collected from the cornea of the eye to facilitate the process of making neural network models . However, m ost of the methods introduced in the past are computationally complicated . In this study, we tried to increase the accessibility of deep neural network models. We did this from the most basic level, i.e. changing and improving the c onvolutional layers. By doing so, we created a new general model that use our new convolutional layer named ArConv layers. Due to the proper functioning of the new layer, the model has suitable complexity for use in mobile phones and perform the task of diagnosing the presence of disease with high accuracy. The final model introduced by us has only 1.3 million parameters and compared to the MobileNetV2 model, which has 2.2 million parameters, after training the model only on the RfMiD data set under the same conditions, results showed that it had better accuracy in the final evaluation on the RfMiD test set. Keywords: Eye disease recognition, Deep convolutional neu ral networks, Machine learning, Computer aided diagnosis, Object detection. Vision is one of the most important senses in humans, according to the evolutionary characteristics of humans; vision is the largest system in brain and occupies 20 - 30% in of the cortex [1] . A s a result, it has a great impact on all aspects of life, including health, the ability to learn and work, help to others and its absence has bad consequences and severely affects people's lives. Eye diseases can cause vision disorders and blindness, and p eople who live in vulnerable communities have less access to medical diagnosis facilities, which will make the problem bigger.