#### Importing your Images
Create your object detection project in Roboflow and start importing your data. We recommend that you import images in a **JPEG, JPG or PNG** format rather than in a PDF format.  
_For this step, a free account on Roboflow is sufficient._  
_Please note that Roboflow does not support the TIFF format._

#### Defining your Annotation Classes
When training an object detection model, you need to determine what are the objects you wish to detect. Each object refers to a **class** associated with a **label**.  
Example of classes: paragraphs, headings, form fields, tables, images, etc.  

#### Using Bounding Boxes
We recommend that you use rectangular bounding boxes rather than polygons to annotate your data because it makes detection easier.  

#### Generating your Dataset
Once your images are annotated according to your defined classes, generate your dataset by clicking on "Add _n_ images to dataset".

#### Exporting the Data
Go to the "Versions" tab and create a new version of your dataset. Disable augmentation if you want your data as they appear in the web application of Roboflow.  
Export your data using the "Export Dataset" button. Choose a **YOLO export format**.

Train a YOLO model using the ultralytics Python library.  
_For more information, please check the [ultralytics documentation](https://docs.ultralytics.com/modes/train/#how-do-i-train-an-object-detection-model-using-ultralytics-yolo11)._

[YALTAi](https://github.com/PonteIneptique/YALTAi) (_You Actually Look Twice At it_) is a Python library that provides an adapter for kraken to use the YOLOv8 object detection routine.  

* Install YALTAi on your machine: ```pip install yaltai```
* In the shell, use the following command to segment the lines with kraken with the zones from the YOLO model: ```yaltai kraken -I "*.jpg" --suffix ".xml" segment --yolo /path/to/yolo/model```

#### Importing your Dataset
At this point, your data should be in an ALTO format (output from YALTAi). Import your data in eScriptorium.  

#### Training the Text Recognition Model
Thanks to the previous steps, your data has already been segmented (regions with YOLO on Roboflow, and lines with kraken via YALTAi). You can either create your own text recognition model with your training data, or fine-tune an existing model.  

#### Evaluating your Model
It is very important to check the performances of your transcription model, in order to finetune it if necessary. Evaluation scores include Word Error Rate (WER) and Character Error Rate (CER).

_For more information, please check Steps 7 and 8 of the "Automatic Text Recognition Roadmap"._

Now that the segmentation and text transcription are complete, you are free to export your data in the format of your choice (ALTO or TXT). 

Sarah Bénière

Thibault Clérice

Juliette Janès

_Please note that this workflow is bound to change once the new version of eScriptorium is released._

This workflow performs Automatic Text Recognition (ATR) with an object detection approach. It uses a YOLO object detection model for segmenting regions, the kraken library for line segmentation, and performs text recognition on the eScriptorium platform. Please note that some prior knowledge in the application of these tools is expected, like using a command-line interface or having a free Roboflow account.

Edit actor

Other URLs relevant to the resource, e.g. ORCID, etc.

URL, ideally landing page, of the resource.

Person/group/body who helped create the resource.

Should be concise and raise interest in the entry. Allows <Link>Markdown</Link> content.

Other URLs relevant to the resource, e.g. code, DOI, etc.

Upload media

URL pointing to info about the access policy.

The activities you can do with the resource. See <Link></Link>.

Name of the conference where paper was given.

Describes the discipline covered by resource. See <Link></Link>.

Locations where the Resource is offered. See <Link></Link>.

URL to helpdesk for incidents & user requests.

Audience targeted by the resource. See <Link></Link>.

A particular published issue of a journal.

Concept or term related to MP entry. See <Link></Link>.

Language(s) in which a resource is available. See <Link></Link>.

Select license pertaining to resource. See <Link></Link>. If license is unknown, use the "Terms of use" property field to describe the usage terms.

Status of the Resource life-cycle. See <Link></Link>.

Mode of use for the resource. See <Link></Link>.

File format of the linked resource. See <Link></Link>.

Page #s of a resource, separate by commas.

Refine the resource's nature and function. See <Link></Link>.

Links to non-MP materials that are relevant.

Info about performance levels provider expected to deliver.

Standard used by the resource. See <Link></Link>.

Technology Readiness Level of the Resource. See <Link></Link>.

Resource user manual and documentation URL.

Version of the tool or service that is linked.

The volume of a journal or multivolume book.

Edit source

Name

If you are having trouble with the sign in process, please <ContactLink>contact the SSH Open Marketplace team</ContactLink>.

If you are having trouble with the sign up process, please <ContactLink>contact the SSH Open Marketplace team</ContactLink>.

Automatic Text Recognition using Object Detection with eScriptorium

Media

Related items(2)

eScriptorium

Roboflow

Workflow steps(6)

1 Annotating Training Data in Roboflow

2 Exporting the Annotated Training Data

3 Training a YOLO Object Detection Model

4 Segmenting Data with YALTAi

5 Text Recognition with eScriptorium

6 Exporting your Data from eScriptorium