Inception-v1-v4 for TF 2.x and Keras 2.4.3

General Introduction

It is an attempt to have the new interpretation of the classical Inception v1~v4 models, Xception v1,Inception ResNet v2 in the new environment. The new interpreted models are used for the thorough study on how the outstanding creators and developers have built the classical network based on the network depth of AlexNet and the small filter size of NIN(Network In Network). It typically stacks the layers for more the complex expression of the image classification.

Characteristics

Address the classical Inception v1~v4 models in TensorFlow 2.3 and Keras 2.4.3. Rebuild the 4 models with the style of linear algebra, including matrix components for both Inception A,B,C and Reduction A,B. In contrast, Inception stem only addresses addition computation.

Operating Environment

Ununtu 18.04

TensorFlow 2.3

Keras 2.4.3

CUDA Toolkit 11.0

cuDNN 8.0.1

CUDA 450.57

Others

Thanks for the all the creators, developers and editors who make the contributions to the widespread application in the image recognition related industries. The classical models work both either the tools ans inheritance that enrich the people's lives.