VisualFlickr

VisualFlickr is a CNN-based tool which can infer emotion and sentiment from Flickr pictures.
It uses Caffe framework, in particular detector provided by MVSO from Columbia University.
VF allows keyword-based search and the analysis of entire profiles from Flickr.
The GUI shows photos, data and pie charts in order to clarify provided information.
All info in VisualFlickr complete white paper.pdf

Requirements

• Caffe deep learning framework installed on your machine
  If you use aptitude package manager, you can run:

  apt install caffe-cpu
  

  or

  apt install caffe-cuda
  

• Python 3.x

• You can install all the necessary packages running:

  pip install -r requirements.txt
  

• ImageTK
  If you use aptitude package manager, you can run

  apt install python3-pil.imagetk
  

Tested

VisualFlickr has been tested on Ubuntu 18.04 in Tilix terminal emulator and in PyCharm IDE.

Run

Convolutional neural network used for this project are available in this Drive folder. You have to download the six directories labelled with the langauge names and put into the net directory. Run demo.py in your Python interpreter.

Tree view of the entire project

VisualFlickr
│   demo.py
│   README.md
│   requirements.txt
│   tags.txt
│   tree.txt
│   users.json
│   
├───image_test
├───modules
├───mvso_scores
│   ├───ANP_emotion_scores
│   └───mvso_sentiment
│           
├───net
│   ├───chinese
│   ├───english
│   ├───french
│   ├───german
│   ├───italian
│   └───spanish
│           
└───settings