The current recipe will cover the set-up for using a pretrained model. We will use TensorFlow to demonstrate the recipe. The current recipe will use VGG16 architecture built using the ImageNet as dataset. The ImageNet is an open source image repository of images used for building image recognition algorithms. The database has more than 10 millions tagged images and more than 1 million images have bounding box to capture objects.
Lot of different deep learning architectures are developed using ImageNet dataset. Once of the popular one is VGG networks are convolution neural networks proposed by Zisserman and Simonyan (2014) and trained over ImageNet data with 1,000 classes. The current recipe will consider VGG16 variant of VGG architecture which is known for it's simplicity. The network uses input of 224 x 224 RGB image. The network utilizes 13 convolution layers with different width x height x depth. The maximum pooling layer is used to reduce volume...