Home

Awesome

MIA (NeurIPS 2019)

Implementation of "Aligning Visual Regions and Textual Concepts for Semantic-Grounded Image Representations" by Fenglin Liu, Yuanxin Liu, Xuancheng Ren, Xiaodong He, and Xu Sun. The paper can be found at [arxiv], [pdf].

Image text

Semantic-Grounded Image Representations (Based on the Bottom-up features)

Coming Soon!

Textual Concepts (Google Drive)

[Pre-trained Models (Google Drive)]

Coming Soon!

Usage

Requirements

This code is written in Python2.7 and requires PyTorch >= 0.4.1

You may take a look at https://github.com/s-gupta/visual-concepts to find how to get the textual concepts of an image by yourself.

Dataset Preparation

Download MSCOCO images and preprocess them

Download the mscoco images from link. You need 2014 training images and 2014 val. images. You should put the train2014/ and val2014/ in the ./data/images/ directory.

Note: We also provided a download bash script to download the mscoco images:

cd data/images/original && bash download_mscoco_images.sh

Now you may need to run resize.py to resize all the images (in both train and val folder) into 256 x 256. You may specify different locations inside resize.py

python resize_images.py

Download MSCOCO captions and preprocess them

You may download the mscoco captions from the official website or use the download bash script provided by us.

cd data && bash download_mscoco_captions.sh

Afterwards, you should create the Karpathy split for training, validation and test.

python KarpathySplit.py

Then you can build the vocabulary by running (Note: You should download the nltk_data to build the vocabulary.)

unzip nltk_data.zip && python build_vocab.py

Download image concepts

Download the Textual Concepts (Google Drive) and put it in the ./data/ directory.

mv image_concepts.json ./data

Start Training

Now you can train the baseline models and the baseline w/ MIA models with:
(Note: We also released the pre-trained models in [Google Drive] (Coming Soon!))

Visual Attention

CUDA_VISIBLE_DEVICES=0,1 python Train.py --basic_model=VisualAttention 
CUDA_VISIBLE_DEVICES=0,1 python Train.py --basic_model=VisualAttention --use_MIA=True --iteration_times=2

Concept Attention

CUDA_VISIBLE_DEVICES=0,1 python Train.py --basic_model=ConceptAttention
CUDA_VISIBLE_DEVICES=0,1 python Train.py --basic_model=ConceptAttention --use_MIA=True --iteration_times=2

Visual Condition

CUDA_VISIBLE_DEVICES=0,1 python Train.py --basic_model=VisualCondition
CUDA_VISIBLE_DEVICES=0,1 python Train.py --basic_model=VisualCondition --use_MIA=True --iteration_times=2

Concept Condition

CUDA_VISIBLE_DEVICES=0,1 python Train.py --basic_model=ConceptCondition
CUDA_VISIBLE_DEVICES=0,1 python Train.py --basic_model=ConceptCondition --use_MIA=True --iteration_times=2

Visual Regional Attention (Coming Soon!)

CUDA_VISIBLE_DEVICES=0,1 python Train.py --basic_model=VisualRegionalAttention
CUDA_VISIBLE_DEVICES=0,1 python Train.py --basic_model=VisualRegionalAttention --use_MIA=True --iteration_times=2

Testing

You can test the trained model with Test.py, but don't forget to download the coco-caption code from link1 or link2 into coco directory.

CUDA_VISIBLE_DEVICES=0 python Test.py  --basic_model=basic_model_name

Note: basic_model_name = (VisualAttention, ConceptAttention, VisualCondition, ConceptCondition, VisualRegionalAttention)

CUDA_VISIBLE_DEVICES=0 python Test.py  --basic_model=basic_model_name --use_MIA=True --iteration_times=2

Reference

If you use this code or our extracted image concepts as part of any published research, please acknowledge the following paper

@inproceedings{Liu2019MIA,
  author    = {Fenglin Liu and
               Yuanxin Liu and
               Xuancheng Ren and
               Xiaodong He and
               Xu Sun},
  title     = {Aligning Visual Regions and Textual Concepts for Semantic-Grounded
               Image Representations},
  booktitle = {NeurIPS},
  pages     = {6847--6857},
  year      = {2019}
}

Acknowledgements

Thanks to Pytorch team for providing Pytorch, COCO team for providing dataset, Tsung-Yi Lin for providing evaluation codes for MS COCO caption generation, Yufeng Ma for providing open source repositories and Torchvision ResNet implementation.

Note

If you have any questions about the code or our paper, please send an email to fenglinliu98@pku.edu.cn