https://github.com/keytoyze/visionts
Code for our paper "VisionTS: Visual Masked Autoencoders Are Free-Lunch Zero-Shot Time Series Forecasters".
https://github.com/keytoyze/visionts
computer-vision deep-learning time-series
Last synced: 2 months ago
JSON representation
Code for our paper "VisionTS: Visual Masked Autoencoders Are Free-Lunch Zero-Shot Time Series Forecasters".
- Host: GitHub
- URL: https://github.com/keytoyze/visionts
- Owner: Keytoyze
- License: mit
- Created: 2024-08-26T17:44:52.000Z (almost 2 years ago)
- Default Branch: main
- Last Pushed: 2025-08-13T08:19:39.000Z (11 months ago)
- Last Synced: 2025-10-27T10:29:54.439Z (8 months ago)
- Topics: computer-vision, deep-learning, time-series
- Language: Python
- Homepage: https://arxiv.org/pdf/2408.17253
- Size: 3.2 MB
- Stars: 255
- Watchers: 4
- Forks: 24
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
- Awesome-MIM - [Code
README
# VisionTS
_Visual Masked Autoencoders Are Free-Lunch Zero-Shot Time Series Forecasters_
[](https://arxiv.org/abs/2408.17253)
[](https://arxiv.org/abs/2508.04379)
[](#-quick-start)
[](https://aihorizonforecast.substack.com/p/visionts-building-high-performance)
[](https://mp.weixin.qq.com/s/vTPkbu5ANYBGO4KYKZfAyg)
πΒ About
| πΒ Quick Start
| πΒ Evaluation
| πΒ Citation
## π What's New
- π₯ Aug 2025: We released [VisionTS++](https://arxiv.org/abs/2508.04379), a SOTA time series foundation model by continual pretraining visual MAE on large-scale time series data, supporting multi-channel forecasting and probablistic forecasting!
- May 2025: Our paper is accepted by ICML 2025!
- Nov 2024: VisionTS achieved the **#1** rank π for zero-shot point forecasting (MASE) on [GIFT-EVAL](https://huggingface.co/spaces/Salesforce/GIFT-Eval) (as of Nov 2024, surpassing Moirai, TimesFM, chronos, etc) β **without any time series training**!
## π About
- We propose **VisionTS**, a time series forecasting (TSF) foundation model building from rich, high-quality *natural images* πΌοΈ.
- This is conceptually different from the existing TSF foundation models (*text-based* π or *time series-based* π), but it shows a comparable or even better performance **without any adaptation on time series data**.
- We reformulate the TSF task as an image reconstruction task, which is further processed by a visual masked autoencoder ([MAE](https://arxiv.org/abs/2111.06377)).
## π Quick Start
We have uploaded our package to PyPI. Please first install [pytorch](https://pytorch.org/get-started/locally/), then running the following command for installing **VisionTS**:
```bash
pip install visionts
```
Then, you can refer to [demo.ipynb](demo.ipynb) about forecasting time series using **VisionTS**, with a clear visualization of the image reconstruction.
## π Evaluation
Our repository is built on [Time-Series-Library](https://github.com/thuml/Time-Series-Library), [MAE](https://github.com/facebookresearch/mae), and [GluonTS](https://github.com/awslabs/gluonts). Please install the dependencies through `requirements.txt` before running the evaluation.
#### Long-Term TSF Benchmarks (Zero-Shot)
We evaluate our methods on 6 long-term TSF benchmarks for zero-shot forecasting. The scripts are under `long_term_tsf/scripts/vision_ts_zeroshot`. Before running, you should first follow the instructions of [Time-Series-Library](https://github.com/thuml/Time-Series-Library) to download datasets into `long_term_tsf/dataset`. Using the following command for reproduction:
```bash
cd long_term_tsf/
bash scripts/vision_ts_zeroshot/$SOME_DATASET.sh
```
#### Monash (Zero-Shot)
We evaluate our methods on 29 Monash TSF benchmarks. You can use the following command for reproduction, where the benchmarks will be automatically downloaded.
```bash
cd eval_gluonts/
bash run_monash.sh
```
> [!IMPORTANT]
> The results in the paper are evaluated based on `python==3.8.18`, `torch==1.7.1`, `torchvision==0.8.2`, and `timm==0.3.2`. Different versions may lead to slightly different performance.
#### PF (Zero-Shot)
We evaluate our methods on 6 long-term TSF benchmarks for zero-shot forecasting. Before running, you should first follow the instructions of [Time-Series-Library](https://github.com/thuml/Time-Series-Library) to download datasets into `long_term_tsf/dataset`, in addition to the following three datasets:
- Walmart: https://www.kaggle.com/competitions/walmart-recruiting-store-sales-forecasting/overview (download to `long_term_tsf/dataset/walmart-recruiting-store-sales-forecasting/train.csv`)
- Istanbul Traffic: https://www.kaggle.com/datasets/leonardo00/istanbul-traffic-index (download to `long_term_tsf/dataset/istanbul-traffic-index/istanbul_traffic.csv`)
- Turkey Power: https://www.kaggle.com/datasets/dharanikra/electrical-power-demand-in-turkey (download to `long_term_tsf/dataset/electrical-power-demand-in-turkey/power Generation and consumption.csv`)
You can use the following command for reproduction.
```bash
cd eval_gluonts/
bash run_pf.sh
```
#### Long-Term TSF Benchmarks (Full-Shot)
We evaluate our methods on 8 long-term TSF benchmarks for full-shot forecasting. The scripts are under `long_term_tsf/scripts/vision_ts_fullshot`. Using the following command for reproduction:
```bash
cd long_term_tsf/
bash scripts/vision_ts_fullshot/$SOME_DATASET.sh
```
## π Citation
```bibtex
@misc{chen2024visionts,
title={VisionTS: Visual Masked Autoencoders Are Free-Lunch Zero-Shot Time Series Forecasters},
author={Mouxiang Chen and Lefei Shen and Zhuo Li and Xiaoyun Joy Wang and Jianling Sun and Chenghao Liu},
year={2024},
eprint={2408.17253},
archivePrefix={arXiv},
url={https://arxiv.org/abs/2408.17253},
}
```
## β Star History
