3-2-1. ResNet-v2-152:
ResNet-v2-152 [16] is a deep neural network architecture with 152 layers, based on the ResNet (Residual Network) design. It is an updated version of the original ResNet, which aims to overcome some of the challenges faced during deep network training. With a large number of parameters, ResNet-v2-152 is considered a very deep network, making it suitable for computer vision tasks such as image classification, object detection, and segmentation. Its depth and the residual connections in its design allow the network to effectively learn hierarchical features and handle high-level abstractions.
3-2-2. Inception-v3 : Inception-v3 [17] is a deep neural network architecture for image classification. It is a variation of the Inception architecture, which was created to tackle challenges associated with training deep networks, such as computational and memory demands. Inception-v3 has been optimized to be computationally efficient while still delivering high accuracy in image classification tasks. The architecture features a modular design, utilizing multiple Inception modules that perform convolution and pooling with various kernel sizes, allowing the network to learn both detailed and broad features. This architecture has been trained on the massive ImageNet dataset and has been shown to achieve top results in different computer vision tasks.