Skip to content
master
Go to file
Code

Latest commit

 

Git stats

Files

Permalink
Failed to load latest commit information.
Type
Name
Latest commit message
Commit time
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

README.md

re-implemented speaker-listener-reinforcer (with ResNet)

URL: https://arxiv.org/pdf/1612.09542.pdf

preprocess

python prepro.py -d refcocog -s google

extract features

python scripts/extract_target_feats.py -d refcocog -s google --batch_size 40 -g 0
python scripts/extract_image_feats.py -d refcocog -s google --batch_size 40 -g 0

training reinforcer

python scripts/train_vlsim.py -d refcocog -s google -g 0 --id slr

if you want to use attention in reinforcer and listener, please include 'attention' in --id.

joint training (speaker, listerner with reinforcer's reward)

python train.py -d refcocog -s google -g 0 --id slr --id2 ver1

evaluation

  • generation
python eval_generation.py -d refcocog -s google -g 1 --id slr --id2 ver1 -split val --batch_size 1
  • comprehension
python eval_comprehension.py -d refcocog -s google -g 1 --id attention -split val --batch_size 1

About

re-implementation of speaker-listener-reinforcer

Resources

Releases

No releases published

Packages

No packages published
You can’t perform that action at this time.