Riksarkivet/trocr-base-handwritten-hist-swe-2

Swedish Lion Libre

An HTR model for historical swedish developed by the Swedish National Archives in collaboration with the Stockholm City Archives, the Finnish National Archives and Jämtlands Fornskriftsällskap. The model is trained on Swedish handwriting dating from ca 1600-1900

Model Details

Model Description

• Developed by: The Swedish National Archives
• Model type: TrOCR base handwritten
• Language(s) (NLP): Historical Swedish handwriting
• License: {{ license | default("[More Information Needed]", true)}}
• Finetuned from model: trocr-base-handwritten

Uses

The model is trained on Swedish running-text handwriting dating from the start of the 17th century to the end of the 19th century. Like most current HTR-models it operates on a text-line level, so it's intended use is within an HTR-pipeline that segments the text into text-lines, which are transcribed by the model.

Direct Use

The model can be used without fine-tuning on all handwriting but performs best on the type of handwriting it was trained on, which is Swedish handwriting from 1600-1900. See below for detailed test and evaluation results.

Downstream Use

The model can be fine-tuned on other types of handwriting, or if you plan to use it to transcribe some specific material that is within it's domain but not included in the training data, for instance if you got a large letter collection dating from the 17th century, it can be fine-tuned on a small amount of manually transcribed in-domain data, say 20-50 letters, and then used to transcribe the entire collection.

Out-of-Scope Use

The model wont work well out-of-the-box for other languages than Swedish, and it wont work well for printed text.

How to Get Started with the Model

Use the code below to get started with the model, but bare in mind that the image has to be a single text-line.

from transformers import TrOCRProcessor, VisionEncoderDecoderModel
from PIL import Image
import requests

img_path = 'path/to/image'
image = Image.open(img_path)

processor = TrOCRProcessor.from_pretrained("microsoft/trocr-base-handwritten")
model = VisionEncoderDecoderModel.from_pretrained('Riksarkivet/trocr-base-handwritten-hist-swe-2')
pixel_values = processor(images=image, return_tensors="pt").pixel_values

generated_ids = model.generate(pixel_values)
generated_text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]

If you want to transcribe entire pages, consider using HTRflow, a package developed by The Swedish National Archives and intended for streamlining large and small scale HTR/OCR-projects. Install the package, write a pipeline config yaml, where you specify the models to use by their huggingface id, add preprocessing or post-processing steps, and then run the pipeline with htrflow pipeline <path/to/yaml> <path/to/image/images>. A .yaml file for an entire pipeline, transcribing full pages, could look like this:

# Demo pipeline for running text

steps:

# Region segmentation
- step: Segmentation
  settings:
    model: yolo
    model_settings:
       model: Riksarkivet/yolov9-regions-1
    generation_settings:
       conf: 0.3
       batch_size: 32

# Line segmentation
- step: Segmentation
  settings:
    model: yolo
    model_settings:
       model: Riksarkivet/yolov9-lines-within-regions-1
    generation_settings:
       conf: 0.3
       batch_size: 16

- step: TextRecognition
  settings:
    model: WordLevelTrocr
    model_settings:
       model: Riksarkivet/trocr-base-handwritten-hist-swe-2
    generation_settings:
       batch_size: 16
       num_beams: 1

- step: ReadingOrderMarginalia
  settings:
    two_page: always

- step: RemoveLowTextConfidencePages
  settings:
    threshold: 0.95

- step: RemoveLowTextConfidenceLines
  settings:
    threshold: 0.95

# Export to Alto and Page XML 
- step: Export
  settings:
    dest: outputs/new_models/alto
    format: alto

- step: Export
  settings:
    dest: outputs/new_models/page
    format: page

# Sets label format to regionX_lineY_wordZ
labels:
  level_labels:
    - region
    - line
    - word
  sep: _
  template: "{label}{number}"

See the documentation for the HTRflow package for further instructions on specific steps and customizations

Training Details

Training Data

We cannot publically release all data the model was trained on, since we ourselves haven't created all the data, but below are links to the datasets we can release publically:

Göteborgs poliskammare 1850-1900
Krigshovrättens domböcker
Svea hovrätt
Bergskollegium
Frihetstidens utskottshandlingar
Carl-Fredrik Påhlmans resejournaler
Trolldomskommissionen
Göta hovrätt
Bergmästaren i Nora
Älvsborgs lösen
Jönköpings rådhusrätt magistrat

Training Procedure

Preprocessing

The text-line polygons were masked out and placed against a white backgroundy, with dimensions decided by the polygon's bounding box.

Training Hyperparameters

See config.json at model repo
training regime: bf16
learning rate: 5e-5
weight decay: 0.01

Evaluation

In-Domain Evaluation Data (Sorted by CER)

These are the character and word error rates on evaluation data taken from the same archives that was included in the training set. Of course the evaluation samples aren't part of the training data. The number of samples included in the training-set give an indication of how the model improves by fine-tuning it on some specific material within the model's range.

Dataset	WER	CER	Train Lines	Eval Lines
krigshovrattens_dombocker_lines	0.0330	0.0075	16,887	1,877
stockholms_stadsarkiv_allmana_barnhuset_1700_lines	0.0647	0.0120	565	142
stockholms_stadsarkiv_blandat_2_1700_lines	0.0807	0.0170	25,024	2,781
goteborgs_poliskammare_fore_1900_lines	0.0800	0.0187	339,297	17,858
stockholms_stadsarkiv_stockholms_domkapitel_1700_lines	0.0948	0.0187	96,409	5,075
stockholms_stadsarkiv_politikollegiet_1700_lines	0.1108	0.0225	120,238	6,329
bergskollegium_relationer_och_skrivelser_lines	0.1056	0.0253	62,201	6,912
stockholms_stadsarkiv_stadens_kamnarsratt_1700_lines	0.1252	0.0278	38,330	4,259
svea_hovratt_lines	0.1484	0.0313	36,884	4,099
stockholms_stadsarkiv_stockholms_domkapitel_1800_lines	0.1400	0.0324	2,070	230
stockholms_stadsarkiv_handelskollegiet_1600_1700_lines	0.1785	0.0350	9,201	1,023
frihetstidens_utskottshandlingar_lines	0.1481	0.0362	13,490	1,499
stockholms_stadsarkiv_kungliga_hovkonsistoriet_1700_lines	0.1541	0.0364	5,753	640
national_archives_finland_court_records_lines	0.1607	0.0368	147,456	7,761
stockholms_stadsarkiv_blandat_1600_1700_lines	0.1505	0.0379	16,137	1,794
stockholms_stadsarkiv_blandat_3_1600_lines	0.1633	0.0400	43,142	4,794
stockholms_stadsarkiv_norra_forstadens_kamnarsratt_1600_1700_lines	0.1755	0.0463	18,474	2,053
carl_fredrik_pahlmans_resejournaler_lines	0.1768	0.0482	7,081	787
stockholms_stadsarkiv_sollentuna_haradsratt_1700_1800_lines	0.1921	0.0505	19,096	2,122
stockholms_stadsarkiv_byggningskollegium_1600_lines	0.2262	0.0514	3,104	345
ra_enstaka_sidor_lines	0.1991	0.0538	5,078	565
trolldomskommissionen_lines	0.2321	0.0600	33,498	3,722
stockholms_stadsarkiv_stockholms_domkapitel_1600_lines	0.2170	0.0607	11,619	1,292
stockholms_stadsarkiv_botkyrka_kyrkoarkiv_1600_1800_lines	0.2548	0.0627	3,617	402
gota_hovratt_lines	0.2450	0.0630	2,421	269
bergmastaren_i_nora_htr_lines	0.2558	0.0709	7,916	880
bergskollegium_advokatfiskalkontoret_lines	0.2906	0.0722	2,411	268
jl_fornsallskap_jamtlands_domsaga_lines	0.2585	0.0732	60,544	6,728
alvsborgs_losen_lines	0.1896	0.0806	5,632	626
jonkopings_radhusratt_och_magistrat_lines	0.2864	0.0853	1,179	131
national_archives_finland_letters_recipes_lines	0.3857	0.1360	651	163

Testing Data

Out-of-Domain Test Data (Sorted by CER)

These are all test-sets taken from archives that we're not at all included in the training data. So these are the results one would expect if one uses this model out-of-the-box on just any running text document within the models time-span. The entire test-suite is available here: test-suite for htr

Dataset	WER	CER	Eval Lines
1792_R0002231_eval_lines	0.1190	0.0250	501
1794-1795_A0068546_eval_lines	0.1503	0.0303	510
1775-1786_A0068551_eval_lines	0.2203	0.0543	525
1841_Z0000017_eval_lines	0.2247	0.0555	470
1690_A0066756_eval_lines	0.2571	0.0611	249
1716_A0017151_eval_lines	0.2517	0.0650	558
1824_H0000743_eval_lines	0.2684	0.0674	260
1699-1700_C0113233_eval_lines	0.2713	0.0691	394
1845-1857_B0000011_eval_lines	0.2546	0.0706	153
1812_A0069332_eval_lines	0.2868	0.0793	69
1659-1674_R0000568_eval_lines	0.3278	0.0886	304
1755-1756_C0112394_eval_lines	0.3440	0.0918	248
1723_H0000374_eval_lines	0.3105	0.1140	378
1887-1892_A0002409_eval_lines	0.3670	0.1297	784
1679_R0002397_eval_lines	0.4768	0.1422	88
1800_C0101725_eval_lines	0.4459	0.1767	37
1871_K0017448_eval_lines	0.4504	0.1945	331
1654_R0001308_eval_lines	0.5200	0.2179	199

Metrics

Character Error Rate (CER)

Character Error Rate (CER) is a metric used to evaluate the performance of a Handwritten Text Recognition (HTR) system by comparing the recognized text to the reference (ground truth) text at the character level.

The CER is calculated using the following formula:

$$CER=\frac{S+D+I}{N}CER\; =\; \backslash frac\{S\; +\; D\; +\; I\}\{N\}$$

Where:

• ( S ) = Number of substitutions (incorrect characters)
• ( D ) = Number of deletions (missing characters)
• ( I ) = Number of insertions (extra characters)
• ( N ) = Total number of characters in the reference text

A lower CER indicates better recognition accuracy.

Word Error Rate (WER)

Word Error Rate (WER) is a metric used to assess the accuracy of an HTR system at the word level by comparing the recognized text to the reference text.

The WER is calculated using the following formula:

$$WER=\frac{S+D+I}{N}WER\; =\; \backslash frac\{S\; +\; D\; +\; I\}\{N\}$$

Where:

• ( S ) = Number of substitutions (incorrect words)
• ( D ) = Number of deletions (missing words)
• ( I ) = Number of insertions (extra words)
• ( N ) = Total number of words in the reference text

Similar to CER, a lower WER indicates better word-level accuracy.

Technical Specifications

Model Architecture

See config.json at model repo

Citation

TrOCR paper