ManagerTower

This repo is the official Pytorch implementation of the paper:

ManagerTower: Aggregating the Insights of Uni-Modal Experts for Vision-Language Representation Learning

Xiao Xu, Bei Li, Chenfei Wu, Shao-Yen Tseng, Anahita Bhiwandiwalla, Shachar Rosenman, Vasudev Lal, Wanxiang Che, Nan Duan.

ACL 2023 (Oral) | Association for Computational Linguistics

Paper | Arxiv | Model | Slides | Video(EN) | Video(CN) | Blog(CN) | Tweet(EN)

Abstract

Two-Tower Vision-Language (VL) models have shown promising improvements on various downstream VL tasks. Although the most advanced work improves performance by building bridges between encoders, it suffers from ineffective layer-by-layer utilization of uni-modal representations and cannot flexibly exploit different levels of uni-modal semantic knowledge. In this work, we propose ManagerTower, a novel VL model architecture that gathers and combines the insights of pre-trained uni-modal experts at different levels. The managers introduced in each cross-modal layer can adaptively aggregate uni-modal semantic knowledge to facilitate more comprehensive cross-modal alignment and fusion. ManagerTower outperforms previous strong baselines both with and without Vision-Language Pre-training (VLP). With only 4M VLP data, ManagerTower achieves superior performances on various downstream VL tasks, especially 79.15% accuracy on VQAv2 Test-Std, 86.56% IR@1 and 95.64% TR@1 on Flickr30K. Code and checkpoints are available at https://github.com/LooperXX/ManagerTower.

Architecture

BridgeTower vs. ManagerTower

Main Results

Visualization

Deployment

• Run setup.sh to set up the environment.
• [Optional] We use wandb to track experiments! Please remember to wandb login and paste your token before running the script.

Dataset Preparation

• We follow ViLT and use pyarrow to serialize the datasets. See here for details.
• For SNLI-VE dataset, we follow here.
• For VG-QA dataset, except the image-text pairs in VG got from here, image meta data, question answers data and coco split information also need to be downloaded.
• The final file structure of datasets are shown in setup.sh.

Checkpoints

We provide the following checkpoints for reproducing our results. You can download them from here.

• Pre-trained checkpoints on 4M data
• Fine-tuned checkpoints for
  • Visual Question Answering on VQAv2
  • Image-Text Retrieval on Flickr30k
  • Visual Entailment on SNLI-VE
  • Visual Reasoning on NLVR$^2$

Pre-training on Image-Text Datasets

# Pre-train ManagerTower Base Model
bash scripts/pre_train.sh

Fine-tuning on Downstream VL Tasks

• VQAv2 Evaluation needs to submit the json file in the logs/ directory to eval.ai evaluation server to get the test-dev and/or test-std scores.

# Base Model on VQAv2 without VLP
bash scripts/ftfs_base_vqa.sh

# Base Model on VQAv2 with VLP
bash scripts/ftfpt_base_vqa.sh

# Base Model on SNLI-VE with VLP
bash scripts/ftfpt_base_snlive.sh

# Base Model on NLVR^2 with VLP
bash scripts/ftfpt_base_nlvr2.sh

# Base Model on IRTR-Flickr30K with VLP (follow ALBEF to use ITC to sample hard negatives for ITM)
bash scripts/ftfpt_base_flickr.sh

Citation

@article{xu2023managertower,
  title={ManagerTower: Aggregating the Insights of Uni-Modal Experts for Vision-Language Representation Learning},
  author={Xu, Xiao and Li, Bei and Wu, Chenfei and Tseng, Shao-Yen and Bhiwandiwalla, Anahita and Rosenman, Shachar and Lal, Vasudev and Che, Wanxiang and Duan, Nan},
  journal={arXiv preprint arXiv:2306.00103},
  year={2023}
}

Acknowledgement

We are highly grateful for the public code of the following papers, our code is partly based on them:

• Main Code: BridgeTower (which is highly based on ViLT and METER)
• Others: CLIP, ALBEF, BLIP