|
No. | DL model | Description | Remark |
|
1. | CNN | It consists of a set of fully connected layers and convolution layers | Require a few input parameters |
2. | DarkNet | Classification model for object detection | Used for real-time object detection |
3. | DNN | It has many hidden nodes compared with the conventional neural network | Performs deep nonlinear analysis |
4. | GoogleNet | It is an improved DL model for image analysis | Used for object detection with a few input parameters |
5. | InceptionResNetV2 | It has a fixable architecture of a CNN | Used for different types of applications |
6. | Inceptionv3 | Third generation of Google's Inception CNN | Used for classifying visual objects for computer vision applications |
7. | LSTM | A type of recurrent neural network (RNN) | Used for dealing with sequences of data |
8. | MobileNetV2 | A lower complexity and model size DL neural network proposed by Google for mobile phone image processing applications | Used for object detection, classification, and semantic segmentation |
9. | NASNet-Large | A CNN modeled to deal with a large scale of image datasets. | Used to classify objects |
10. | ResNet34 | A CNN architecture but with shortcuts and bottleneck block mechanisms between layers to speed up solving problems. | Used for deep real-time analysis |
11. | ResNet50 | A type of CNN that performs deeper analysis to solve complex problems | The deeper analysis might degrade the accuracy of the network |
12. | SAE | A multilayer neural network with a stacked autoencoder | Used for datasets with a small dimension of features. |
13. | VGG16 | A CNN with multiple 3 × 3 kernel-sized filters in the convolutional layers | Used for recognition tasks of a large-scale number of images dataset |
14. | VGG19 | A CNN with multiple 3 × 3 kernel-sized filters in the convolutional layers with additional layers than the VGG16 | Used for recognition tasks of a large-scale number of images dataset |
15. | Xception | An improved version of the Inception family of CNN | Used for classifying visual objects for computer vision applications with a slightly higher accuracy |
|