Using TensorFlow to classify hotdogs! – Above Intelligent (AI)

In the very popular show Silicon Valley, one of my favorite characters Jian Yang creates a Deep learning application which accurately predicts if a food item is a hot dog or not, pretty funny stuff, So I thought of using google's open source TensorFlow library to create my very own Hot dog classification program. So without further chit chat let's start classifying some hotdogs. For this project I've used: First we have to set up docker, since I'm using debian it was pretty straight forward. Let's launch that docker instance shall we: Since we're classifying if an article is a hotdog or not, we're going to need 1 folder and 2 subfolders, the main folder is going to be called images and within that folder we will create 2 subfolders named hotdogs and random. Now we need around 100 images of hotdogs in the hotdogs folder and 100 random images of things that aren't hotdogs, I used Fatkun image downloader from chrome extension store for this purpose, but a point to be noted is that TensorFlow only handles JPEG images and using PNG's can run you into a lot of trouble, after you're done populating the directories with images of hotdogs and random things in their respective folders we have to download and retrain the Inception V3 net by google, since we don't have the required time nor resources to train our own CNN.