image-captioning-for-celebrities

[updating]

• image-captioning-for-celebrities
  • Overall architecture:
  • Image captioning part:
  • Face recognition part:
  • People noun phrase chunk matching:
  • Datasets
  • Dependencies
  • Usage:

Overall architecture:

Image captioning part:

Face recognition part:

We utilize the mtcnn module and the pretrained Inception_v1 in facenet-pytorch

People noun phrase chunk matching:

Datasets

Download the following datasets and add them to relative paths:

1. Flickr 8k
2. Flickr 30k

Dependencies

torch=1.10.1
python=3.9.7
numpy=1.21.5
matplotlib=3.5.1
torchvision=0.11.2 
spacy=3.2.1
nltk=3.7

Usage:

model training is in main.py. For a glance of the performance, see example.ipynb.