Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
UniTable: Towards a Unified Table Foundation Model
| Date | Stars |
|---|---|
| 2026-07-31 | 534 |
| 2026-08-04 | 534 |
| 2026-08-06 | 534 |
Today
— stars today
This week
— stars this week
This month
— stars this month
Momentum
0.0
growth rate 0.00%/day
# UniTable: Towards a Unified Table Foundation Model <p align="center"><img src="./website/unitable-demo.gif" alt="Demo" width="700"/></p> 1. 📈 [High-Performance Transformers for Table Structure Recognition Need Early Convolutions](https://arxiv.org/abs/2311.05565). ShengYun Peng, Seongmin Lee, Xiaojing Wang, Rajarajeswari Balasubramaniyan, Duen Horng Chau. In *NeurIPS Second Table Representation Learning Workshop*, 2023. (Oral) 2. 🚀 [Self-Supervised Pretraining for Table Structure Recognition Transformer](https://arxiv.org/abs/2402.15578). ShengYun Peng, Seongmin Lee, Xiaojing Wang, Rajarajeswari Balasubramaniyan, Duen Horng Chau. In *AAAI Scientific Document Understanding Workshop*, 2024. (Oral) 3. 🆕 [UniTable: Towards a Unified Framework for Table Structure Recognition via Self-Supervised Pretraining](https://arxiv.org/abs/2403.04822). ShengYun Peng, Seongmin Lee, Xiaojing Wang, Rajarajeswari Balasubramaniyan, Duen Horng Chau. ArXiv, 2024. Tables convey factual and quantitative data with implicit conventions created by humans that are often challenging for machines to parse. Prior work on table recognition (TR) has mainly centered around complex task-specific combinations of available inputs and tools. We present UniTable, a training framework that unifies training paradigm, training objective, and model architecture of TR. Its training paradigm combines the simplicity of purely pixel-level inputs with the effectiveness and scalability empowered by self-supervised pretraining (SSP) from diverse unannotated tabular images. Our framework unifies the training of all three TR tasks — extracting table structure, cell content, and cell bounding box (bbox) — into a unified task-agnostic training objective: language modeling. Extensive quantitative and qualitative analyses highlight UniTable’s state-of-the-art (SOTA) performance on four of the largest TR datasets. To promote reproducible research, enhance transparency, and SOTA innovations, we have released the first-of-its-kind [Jupyter Notebook](./notebooks/full_pipeline.ipynb) of the entire inference pipeline, fine-tuned across multiple TR datasets, supporting all three TR tasks. > This repo includes code for linear projection Transformers. For convolutional stem (early convolution) Transformers, please check out our [tsr-convstem repo](https://github.com/poloclub/tsr-convstem). # News `Apr. 2024` - You can fully digitalize your own tabular image in our [Jupyter Notebook](./notebooks/full_pipeline.ipynb). `Apr. 2024` - UniTable v1.0.0 is now online with model weights available at [HuggingFace](https://huggingface.co/poloclub/UniTable/tree/main). `Feb. 2024` - We presented "Self-Supervised Pretraining" paper at AAAI'24. `Jan. 2024` - "Self-Supervised Pretraining" paper was selected as [oral](https://sites.google.com/view/sdu-aaai24/schedule?authuser=0). `Dec. 2023` - "Self-Supervised Pretraining" paper was accepted by [AAAI'24 Scientific Document Understanding Workshop](https://sites.google.com/view/sdu-aaai24/schedule?authuser=0). `Dec. 2023` - We presented "Early Convolutions" paper at [NeurIPS'23](https://x.com/RealAnthonyPeng/status/1735715161476866135?s=20). `Oct. 2023` - "Early Convolutions" paper was selected as [oral](https://table-representation-learning.github.io/#accepted-papers). `Oct. 2023` - "Early Convolutions" paper was accepted by [NeurIPS'23 Table Representation Learning Workshop](https://table-representation-learning.github.io/). # Quick Start 1. Set up virtual environment (unitable) by running `make .done_venv` in your terminal. 2. Download all the model weights from [HuggingFace](https://huggingface.co/poloclub/UniTable/tree/main) by running `make .done_download_weights` in your terminal. 3. Try out our demo [Jupyter Notebook](./notebooks/full_pipeline.ipynb) with your own tabular image! Remember to select "unitable" as your notebook kernel. # Training Our code is driven by [Makefile targets](https://www.gnu.org/software/make/manual/mak
Excerpt of 8,334 characters
Read on GitHub
5
3
Would you bet a product on this? Bounded 0–100 and slow moving.
matched fp:d19a4787c786c1ad, llm:Repository title/description: 'UniTable: Towards a Unified Table Foundation Model' (Jupyter Notebook).
matched fp:d19a4787c786c1ad, llm:Repository title/description: 'UniTable: Towards a Unified Table Foundation Model' (Jupyter Notebook).
matched fp:d19a4787c786c1ad, llm:Repository title/description: 'UniTable: Towards a Unified Table Foundation Model' (Jupyter Notebook).