GitPlanet

Cheap and reliable Node.js hosting starts at $3/month, and $1/month static HTML hosting

Created with love in Canada, visit hostnodejs.com today

Feel like to post an Ad? Learn Details
All Projects → eric-xw → Arel

eric-xw / Arel

Licence: mit
Code for the ACL paper "No Metrics Are Perfect: Adversarial Reward Learning for Visual Storytelling"

Programming Languages

python
139335 projects - #7 most used programming language

Labels

rladversarial-learning

Projects that are alternatives of or similar to Arel

Nlg Rl
Accelerated Reinforcement Learning for Sentence Generation by Vocabulary Prediction
Stars: ✭ 59 (-52.42%)
Mutual labels:  rl
Virtual Adversarial Training
Pytorch implementation of Virtual Adversarial Training
Stars: ✭ 94 (-24.19%)
Mutual labels:  adversarial-learning
Pytorch Adversarial box
PyTorch library for adversarial attack and training
Stars: ✭ 104 (-16.13%)
Mutual labels:  adversarial-learning
Trading Gym
A Trading environment base on Gym
Stars: ✭ 71 (-42.74%)
Mutual labels:  rl
Ai Paper Drawer
人工智能论文关键点集结。This project aims to collect key points of AI papers.
Stars: ✭ 82 (-33.87%)
Mutual labels:  rl
Exploration By Disagreement
[ICML 2019] TensorFlow Code for Self-Supervised Exploration via Disagreement
Stars: ✭ 99 (-20.16%)
Mutual labels:  rl
Handwriting recogition using adversarial learning
[CVPR 2019] "Handwriting Recognition in Low-resource Scripts using Adversarial Learning ”, IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), 2019.
Stars: ✭ 52 (-58.06%)
Mutual labels:  adversarial-learning
Ros2learn
ROS 2 enabled Machine Learning algorithms
Stars: ✭ 119 (-4.03%)
Mutual labels:  rl
Pose Adv Aug
Code for "Jointly Optimize Data Augmentation and Network Training: Adversarial Data Augmentation in Human Pose Estimation" (CVPR 2018)
Stars: ✭ 83 (-33.06%)
Mutual labels:  adversarial-learning
Aws Robomaker Sample Application Deepracer
Use AWS RoboMaker and demonstrate running a simulation which trains a reinforcement learning (RL) model to drive a car around a track
Stars: ✭ 105 (-15.32%)
Mutual labels:  rl
Noreward Rl
[ICML 2017] TensorFlow code for Curiosity-driven Exploration for Deep Reinforcement Learning
Stars: ✭ 1,176 (+848.39%)
Mutual labels:  rl
Rlenv.directory
Explore and find reinforcement learning environments in a list of 150+ open source environments.
Stars: ✭ 79 (-36.29%)
Mutual labels:  rl
Dopamine
Dopamine is a research framework for fast prototyping of reinforcement learning algorithms.
Stars: ✭ 9,681 (+7707.26%)
Mutual labels:  rl
Ali Pytorch
PyTorch implementation of Adversarially Learned Inference (BiGAN).
Stars: ✭ 61 (-50.81%)
Mutual labels:  adversarial-learning
Gpnd
Generative Probabilistic Novelty Detection with Adversarial Autoencoders
Stars: ✭ 112 (-9.68%)
Mutual labels:  adversarial-learning
Adversarial Explainable Ai
💡 A curated list of adversarial attacks on model explanations
Stars: ✭ 56 (-54.84%)
Mutual labels:  adversarial-learning
Lggan
[CVPR 2020] Local Class-Specific and Global Image-Level Generative Adversarial Networks for Semantic-Guided Scene Generation
Stars: ✭ 97 (-21.77%)
Mutual labels:  adversarial-learning
Pytorch Rl
Tutorials for reinforcement learning in PyTorch and Gym by implementing a few of the popular algorithms. [IN PROGRESS]
Stars: ✭ 121 (-2.42%)
Mutual labels:  rl
Stable Baselines
Mirror of Stable-Baselines: a fork of OpenAI Baselines, implementations of reinforcement learning algorithms
Stars: ✭ 115 (-7.26%)
Mutual labels:  rl
Zerospeech Tts Without T
A Pytorch implementation for the ZeroSpeech 2019 challenge.
Stars: ✭ 100 (-19.35%)
Mutual labels:  adversarial-learning
View All Similar Projects ➔

No Metrics Are Perfect: Adversarial REward Learning for Visual Storytelling

This repo is the implementation of our paper "No Metrics Are Perfect: Adversarial Reward Learning for Visual Storytelling", which also provides a codebase for the task of visual storytelling.

In the AREL paper, we not only introduce a novel adversarial reward learning algorithm to generate more human-like stories given image sequences, but also empirically analyze the limitations of the automatic metrics for story evaluation. For more details, please check the latest version of the paper: https://arxiv.org/abs/1804.09160.

Prerequisites

  • Python 2.7
  • PyTorch 0.3
  • TensorFlow (optional, only using the fantastic tensorboard)
  • cuda & cudnn

Usage

1. Setup

Clone this github repository recursively:

git clone --recursive https://github.com/eric-xw/AREL.git ./

Download the preprocessed ResNet-152 features here and unzip it into DATADIR/resnet_features.

2. Supervised Learning

We use cross entropy loss to warm start the model first:

python train.py --id XE --data_dir DATADIR --start_rl -1

Check the file opt.py for more options, where you can play with some other settings.

3. AREL Learning

To train an AREL model, run

python train_AREL.py --id AREL --start_from_model PRETRAINED_MODEL

Note that PRETRAINED_MODEL can be data/save/XE/model.pth or some other saved models. Check opt.py for more information.

4. Monitor your training

TensorBoard is used to monitor the training process. Suppose you set the option checkpoint_path as data/save, then run

tensorboard --logdir data/save/tensorboard

And then open your browser and go to [IP address]:6006 (the default port for tensorboard is 6006).

5. Testing

To test the model's performance, run

python train.py --option test --beam_size 3 --start_from_model data/save/XE/model.pth

or

python train_AREL.py --option test --beam_size 3 --start_from_model data/save/AREL/model.pth

Reproducing our results

We uploaded our checkpoints and meta files to the IRL-ini-iter100-*. Please load the model from these folders by running

python train.py --option test --beam_size 3 --start_from_model [best_model_path]

If you find this code useful, please cite the paper

@InProceedings{xwang-2018-AREL,
  author = 	"Wang, Xin and Chen, Wenhu and Wang, Yuan-Fang and Wang, William Yang",
  title = 	"No Metrics Are Perfect: Adversarial Reward Learning for Visual Storytelling",
  booktitle = 	"Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
  year = 	"2018",
  publisher = 	"Association for Computational Linguistics",
  pages = 	"899--909",
  location = 	"Melbourne, Australia",
  url = 	"http://aclweb.org/anthology/P18-1083"
}

Acknowledgement

Note that the project description data, including the texts, logos, images, and/or trademarks, for each open source project belongs to its rightful owner. If you wish to add or remove any projects, please contact us at [email protected].