NeuralTalk

Forked from NeuralTalk
Add a single python function to generate sentence from image.
An original Readme.md is placed at neuraltalk/Readme.md

Prerequisites

Caffe and Caffe's python interface installed on your environment

Setup

Setup option 1 (manual)

Install NeuralTalk using pip sudo pip install git+https://github.com/Thanabhat/neuraltalk.git
Download VGG_ILSVRC_16_layers.caffemodel from http://www.robots.ox.ac.uk/~vgg/research/very_deep/ and place at python_features/VGG_ILSVRC_16_layers.caffemodel
Download coco_cnn_lstm_v2.zip from http://cs.stanford.edu/people/karpathy/neuraltalk/ , extract and place at model/model_checkpoint_coco_visionlab43.stanford.edu_lstm_11.14.p

Setup option 2 (script)

You may need to change the path '/usr/local/lib/python2.7/dist-packages/' to match your environment.

# install
sudo pip install git+https://github.com/Thanabhat/neuraltalk.git
sudo wget -O /usr/local/lib/python2.7/dist-packages/neuraltalk/python_features/VGG_ILSVRC_16_layers.caffemodel "http://www.robots.ox.ac.uk/~vgg/software/very_deep/caffe/VGG_ILSVRC_16_layers.caffemodel"
sudo wget -O /home/coco_cnn_lstm_v2.zip "http://cs.stanford.edu/people/karpathy/neuraltalk/coco_cnn_lstm_v2.zip" && sudo unzip /home/coco_cnn_lstm_v2.zip -d /usr/local/lib/python2.7/dist-packages/neuraltalk/model && sudo rm /home/coco_cnn_lstm_v2.zip
# test
sudo python -m neuraltalk.caption_generator

Setup option 3 (docker)

https://hub.docker.com/r/thanabhat/docker-neuraltalk/

Example

Write this python code, place some img.jpg, run and get the sentence!

from neuraltalk.caption_generator import CaptionGenerator

caption_generator = CaptionGenerator()
caption = caption_generator.get_caption('img.jpg')
print(caption)