Convolutional Neural Network / CNN
The type of nervous network that learns by scanning the traces of raditional images with filters.
Evilli nerve network (Convolutional Neural Network, CNN) is a neuro network architecture inspired by the biological vision system, specifically designed to learn the images in image and spatial data. The basic structure stone is to detect the local images in each region, called small size "filter" or "core". The filters in the first layers often learn simple edge, corner and color transitions, the next layers as the network is deepened, combine these simple features of texture, shape and finally can recognize complex object parts such as eye, wheel or face. Pooling (pooling) layers reduce the calculation load by reducing the image size and ensure that the model is resistant to small location slips.
The most important advantage of CNN is "parameter sharing": since the same filter image is used in every region, the model can be trained with much less parameters compared to a fully connected network, and no matter where an object is displayed. These features have made CNNs a standard approach in tasks such as image classification, object detection and medical image analysis for many years; architectures such as AlexNet and ResNet are among kilometers stones of the area. In recent years, attention-based architectures such as Vision Transformer have captured the performance of CNNs in some tasks, CNNs are widely used due to their efficiency and good understood behaviour.
The most important advantage of CNN is "parameter sharing": since the same filter image is used in every region, the model can be trained with much less parameters compared to a fully connected network, and no matter where an object is displayed. These features have made CNNs a standard approach in tasks such as image classification, object detection and medical image analysis for many years; architectures such as AlexNet and ResNet are among kilometers stones of the area. In recent years, attention-based architectures such as Vision Transformer have captured the performance of CNNs in some tasks, CNNs are widely used due to their efficiency and good understood behaviour.