Predicting image classes using Inception V3

Inception V3 is one of the most widely used neural networks, both as an image classifier and a feature extractor. It has over 75% - 78% accuracy on an ImageNet dataset and is proven to be fast and precise. 

Setting up the Inception V3 environment

To start working with Inception V3, you first need to download model files from the MXNet Model Zoo. Use the following links:

• URL: http://data.dmlc.ml/models/imagenet/
• ZIP: inception-v3.tar.gz

After you have downloaded the archive, please extract it to the weights/inception-v3 folder. 

Loading the network

Loading a pre-trained model is similar to loading your model from a checkpoint. We will use the mx.load_checkpoint function to instantiate the model from a checkpoint and also load the synset file containing the class names. This is done with the following code:

using Images, MXNet

const MODEL_NAME = "weights/inception-v3/Inception-7"
const MODEL_CLASS_NAMES = "weights/inception-v3/synset.txt"

nnet = mx.load_checkpoint...