mr-Blip

KO | EN

gitlite — search

#typescript #ai-agents #ai #dsh-plugin #deepseek-harness #open-source #claude-code #codex #cli #developer-tools #react #windows

★ 96 Open GitHub ↗

Official Implementation of "Chrono: A Simple Blueprint for Representing Time in MLLMs"

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# Chrono: A Simple Blueprint for Representing Time in MLLMs

* Authors: [Boris Meinardus](https://sudo-boris.github.io/), [Hector Garcia Rodriguez](https://hector.gr/), [Anil Batra](https://anilbatra2185.github.io/), [Anna Rohrbach](https://anna-rohrbach.net/), [Marcus Rohrbach](https://rohrbach.vision/)
* Paper: [arxiv](http://arxiv.org/abs/2406.18113)

The recent success of Large Language Models (LLMs) has prompted the extension to the multimodal domain developing image-text Multimodal LLMs (MLLMs) and then video-text models. In this work, we investigate the challenge of contextual and temporal comprehension in video-language models by exploring the task of temporal localization in videos. To address this problem, prior works have developed complex task-specific architectures, novel modules to embed time into MLLMs, or leveraged additional input signals such as video transcripts to best encode contextual and temporal information. Interestingly, we find that most of these efforts are surpassed by a much simpler design. We introduce Chrono, a universal sequence blueprint that can be applied to an image-text pretrained MLLM. Through extensive ablations across different MLLM architectures, finetuning and zero-shot settings, and different datasets, we achieve a new SOTA in moment retrieval on the most widely used benchmarks Charades-STA, QVHighlights, ActivityNet Captions, and grounded video question answering on NeXT-GQA.

# Code structure

```bash

# data & data preprocessing
./mr_BLIP_data

# pretrained checkpoints
./mr_BLIP_checkpoints

# mr_BLIP code
./lavis/

# running scripts for mr_BLIP training and inference
./run_scripts

```

# Setup

## Install Dependencies

1. (Optional) Creating conda environment

```bash
conda create -n mrBlip python=3.8
conda activate mrBlip
```

2. build from source

```bash
pip install -r requirements.txt
```

## Download Pretrained Models

We train Mr. BLIP on QVHighlights, Charades-STA, and ActivityNet Captions and provide the checkpoints.
Download the [checkpoints](https://drive.google.com/drive/folders/1AR-rdUillx0fy7KS4zbEuswFMl7qR9Gj?usp=sharing) and put them under /mr_BLIP_checkpoints.

# Dataset Preparation

We test our model on:

* [Charades-STA](https://github.com/jiyanggao/TALL)
  * [Charades (Videos)](https://prior.allenai.org/projects/charades)

* [QVHighlights](https://github.com/jayleicn/moment_detr)

* [ActivityNet Captions](https://cs.stanford.edu/people/ranjaykrishna/densevid/)

Please download original MR data and preprocess them via our [scripts](mr_BLIP_data/data_preprocess.ipynb).

# Training and Inference

We provide Mr. BLIP training and inference script examples as follows.

And please refer to [dataset page](lavis/configs/datasets/) to customize your data path.

You might want to update the [config files](lavis/projects/mr_BLIP/train/) for the respective runs to fit on your machine. They are currently set to run on 8 A100-80GB GPUs. You can simply reduce the batch size, reduce the number of frames, or apply a frame level embeddings aggregation (32 frame tokens -> 1 token) to fit on a smaller GPU.

## 1) QVH Finetuning

```bash
sh run_scripts/mr_BLIP/train/qvh.sh
```

## 2) Charades-STA Finetuning

```bash
sh run_scripts/mr_BLIP/train/charades.sh
```

## 3) ANet Captions Finetuning

```bash
sh run_scripts/mr_BLIP/train/anet.sh
```

## 4) QVH Evaluation

Should roughly return:
|               | R1@0.5        | R1@0.7   | mIoU       | mAP@0.5  | mAP@0.75  |
| ------------- |:-------------:| :-----:  | :--------: | :-----:  | --------: |
| *Mr. BLIP*    | **76.16**     | **62.63**| **70.32**  | **68.50**| **55.06** |

```bash
sh run_scripts/mr_BLIP/eval/qvh.sh
```

## 5) Charades-STA Evaluation

Should roughly return:
|               | R1@0.5        | R1@0.7   | mIoU      |
| ------------- |:-------------:| :-----:  | --------: |
| *Mr. BLIP*    | **69.31**     | **49.29**| **58.63** |

```bash
sh run_scripts/mr_BLIP/eval/charades.sh
```

## 6) ANet Captions Evaluation

Should roughly return:
|               | R1@0.5        | R1@0.7   | mIoU      |
| ------------- |:-------------:| :-----:  | --------: |
| *Mr. BLIP*    | **53.79**     | **35.47**| **51.52** |

```bash
sh run_scripts/mr_BLIP/eval/anet.sh
```

# Acknowledgments

We thank the developers of [LAVIS](https://github.com/salesforce/LAVIS) and [BLIP-2](https://github.com/salesforce/LAVIS/tree/main/projects/blip2) for their public code release.

# Reference

Please cite our paper if you use our models in your works:

```bibtex
@article{meinardus2025chronosimpleblueprintrepresenting,
      title={Chrono: A Simple Blueprint for Representing Time in MLLMs}, 
      author={Boris Meinardus and Hector Garcia Rodriguez and Anil Batra and Anna Rohrbach and Marcus Rohrbach},
      year={2025},
      eprint={2406.18113},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2406.18113}, 
}
```

mr-Blip

Explore Similar Repositories

Was this content helpful?

Recent Feedback

Download README