Image Classifier using tensorflow

This is the code for 'Image Classifier in TensorFlow' by Rishab Sharma (rishab-sharma).

Requirements

• docker

Usage

You just need to make a "classifier" directory with a directory "data" inside it with all your images For example

 [any_path]/my_own_classifier/
 [any_path]/my_own_classifier/data
 [any_path]/my_own_classifier/data/car
 [any_path]/my_own_classifier/data/moto
 [any_path]/my_own_classifier/data/bus

and then put your image on it. This "classifier" directory will have your samples but also trained classifier after execution of "train.sh".

Train process

Just type

 ./train.sh [any_path]/my_own_classifier

And it will do anything for you !

Guess process

Just type for a single guess

 ./guess.sh [any_path]/my_own_classifier /yourfile.jpg

To guess an entire directory

./guessDir.sh [any_path]/classifier [any_path]/srcDir [any_path]/destDir

Example of result

# ./guess.sh /synced/tensor-lib/moto-classifier/ /synced/imagesToTest/moto21.jpg
moto (score = 0.88331)
car (score = 0.11669)

Use an absolute file path for classifier and images because the script dos not support relative path (volume mounting)