OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Tsr Convstem

Tsr Convstem is a Python research repository implementing transformer-based table-structure recognition from images, with a lightweight convolutional stem as the visual encoder and a textual decoder that generates table-representing tokens.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Tsr Convstem — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when you need a research codebase for experimenting with transformer-based table recognition and a documented dataset-to-evaluation workflow.

Who it is for

  • Researchers studying table-structure recognition and visual encoders.
  • Developers implementing or evaluating workflows that convert table images into machine-readable structures.

Documented uses

  • Train a table-structure recognition model using the documented PubTabNet workflow.
  • Test a configured model and compute its TEDS score.
  • Compare visual-encoder and model configurations described in the Makefile.
Before you choose

Requirements & limits

The public GitHub repository identifies its license as MIT.

  • The documented setup expects PubTabNet and a dataset path configured in the Makefile.
  • Training and evaluation are invoked through Makefile targets, with experiment configurations also defined there.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Set up the documented PubTabNet workflow

Prepare the repository for its documented training and evaluation steps.

Before you start

  • The Tsr Convstem repository
  • The PubTabNet dataset
  • A path to PubTabNet to enter in the Makefile

Steps

  1. Obtain the PubTabNet dataset using the dataset instructions linked from the repository README.

  2. Set the Makefile's "pubtabnet_dir" to the dataset path.

  3. Set up the virtual environment with the command below.

    Documented command
    make .venv_done

Expected outcome

The dataset path is configured and the README's virtual-environment setup target has been run; the repository's training and evaluation instructions are the next documented stage.

Guide limitations

  • The README does not specify environment versions in the cited instructions.
  • This setup guide does not cover training or evaluation.
Documentation 1 ↗
Source-based practical guide

Train and evaluate the documented ResNet-18 configuration

Run the README's example training and TEDS evaluation targets for the ResNet-18 experiment.

Before you start

  • The repository setup and PubTabNet path configuration described in its README

Steps

  1. Run the command below to train the documented visual-encoder instance with ResNet-18.

    Documented command
    make experiments/r18_e2_d4_adamw/.done_train_structure

  2. Run the command below to test and compute the TEDS score.

    Documented command
    make experiments/r18_e2_d4_adamw/.done_teds_structure

Expected outcome

The documented targets specify a training stage followed by testing and TEDS-score computation.

Guide limitations

  • The README does not state expected runtime, hardware requirements, or benchmark results for a particular run.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
14 October 2026
Original website title
poloclub/tsr-convstem · Official homepage title · retained in its source language
How we review sourcesSuggest a correction