InceptionTube

A library for downloading and classifying Youtube videos with Inception V3 on Keras/Tensorflow

Installation

Create a new folder and clone the repository

mkdir InceptionTube
cd InceptionTube
git clone git remote add origin https://github.com/gabriele6/InceptionTube.git
git pull origin master

Install the environment by using the environment.yml file

$ conda env create -f environment.yml

How to use

From your terminal, get into the Conda environment created:

$ source activate test.py

The library contains an instantiable class. Create a my_script.py file, put this in it

from youtube_inception import YoutubeInception 
yt = YoutubeInception()

Now you can run every method in the YoutubeInception class by calling it from the instantiated Object

result = yt.youtubeQuery("funny cats video")

NOTE: by default, you need videos/ and screens/ folders in the same directory.

$ mkdir ./videos  
$ mkdir ./screens

You can change the directories by using the setVideoPath and setScreensPath methods.

Non-trivial application

Look at the main.py file. It calls a Youtube query and analyzes videos until it finds the first n videos containing the requested category.
Usage:

python main.py "query" "category" n   

ex:

$ python main.py "surfing sea lion" "sea_lion" 3

The final output is a list of n videos containing the requested category.

Please note, it may take a while to execute, depending on your hardware's capabilities and download speed.