Daniel van Strien
@danielvanstrien.bsky.social
Machine Learning Librarian at @hf.co
Made some improvements to the OCR scripts onboarding in my uv-scripts collection.
First run is now a small example: seven scanned NASA pages in, Markdown out, one HF jobs command. It also works on your own images or a folder of PDFs. You can pick from 28 OCR models.
huggingface.co/datasets/uv-...
First run is now a small example: seven scanned NASA pages in, Markdown out, one HF jobs command. It also works on your own images or a folder of PDFs. You can pick from 28 OCR models.
huggingface.co/datasets/uv-...
September 16, 2026 at 3:09 PM
Made some improvements to the OCR scripts onboarding in my uv-scripts collection.
First run is now a small example: seven scanned NASA pages in, Markdown out, one HF jobs command. It also works on your own images or a folder of PDFs. You can pick from 28 OCR models.
huggingface.co/datasets/uv-...
First run is now a small example: seven scanned NASA pages in, Markdown out, one HF jobs command. It also works on your own images or a folder of PDFs. You can pick from 28 OCR models.
huggingface.co/datasets/uv-...
Reposted by Daniel van Strien
Need a historical illustration?
Search 1.49 million images from British Library books and Britannica (1500s–1920s).
Find similar images, download full-resolution images or transparent PNG cutouts, and explore the sources.
huggingface.co/spaces/davan...
Search 1.49 million images from British Library books and Britannica (1500s–1920s).
Find similar images, download full-resolution images or transparent PNG cutouts, and explore the sources.
huggingface.co/spaces/davan...
September 11, 2026 at 1:42 PM
Need a historical illustration?
Search 1.49 million images from British Library books and Britannica (1500s–1920s).
Find similar images, download full-resolution images or transparent PNG cutouts, and explore the sources.
huggingface.co/spaces/davan...
Search 1.49 million images from British Library books and Britannica (1500s–1920s).
Find similar images, download full-resolution images or transparent PNG cutouts, and explore the sources.
huggingface.co/spaces/davan...
Need a historical illustration?
Search 1.49 million images from British Library books and Britannica (1500s–1920s).
Find similar images, download full-resolution images or transparent PNG cutouts, and explore the sources.
huggingface.co/spaces/davan...
Search 1.49 million images from British Library books and Britannica (1500s–1920s).
Find similar images, download full-resolution images or transparent PNG cutouts, and explore the sources.
huggingface.co/spaces/davan...
September 11, 2026 at 1:42 PM
Need a historical illustration?
Search 1.49 million images from British Library books and Britannica (1500s–1920s).
Find similar images, download full-resolution images or transparent PNG cutouts, and explore the sources.
huggingface.co/spaces/davan...
Search 1.49 million images from British Library books and Britannica (1500s–1920s).
Find similar images, download full-resolution images or transparent PNG cutouts, and explore the sources.
huggingface.co/spaces/davan...
Used Astra + Jobs to see how well this model performs on the full olmOCR-bench
Unsurprisingly, it doesn’t do brilliantly overall: 36.8%
But for a ~16M-parameter recogniser, I think 74.4% on long/tiny text and 57.9% on multi-column pages are pretty interesting.
Unsurprisingly, it doesn’t do brilliantly overall: 36.8%
But for a ~16M-parameter recogniser, I think 74.4% on long/tiny text and 57.9% on multi-column pages are pretty interesting.
A 15.9M-parameter specialist OCR model outperforms many VLMs hundreds of times larger.
On a 2,165-page historical eval dataset, Kraken PP-OCRv6 ranks #4 on reading CER and #1 when long-s, ligatures and case are preserved.
huggingface.co/spaces/fineb...
On a 2,165-page historical eval dataset, Kraken PP-OCRv6 ranks #4 on reading CER and #1 when long-s, ligatures and case are preserved.
huggingface.co/spaces/fineb...
September 10, 2026 at 9:29 AM
Used Astra + Jobs to see how well this model performs on the full olmOCR-bench
Unsurprisingly, it doesn’t do brilliantly overall: 36.8%
But for a ~16M-parameter recogniser, I think 74.4% on long/tiny text and 57.9% on multi-column pages are pretty interesting.
Unsurprisingly, it doesn’t do brilliantly overall: 36.8%
But for a ~16M-parameter recogniser, I think 74.4% on long/tiny text and 57.9% on multi-column pages are pretty interesting.
Reposted by Daniel van Strien
DeepSeek v4.1 Flash is just another level 🤯
huggingface.co/deepseek-ai/...
- Asymmetric Causal-Encoder-Decoder: 550B MoE, input 8B / output 16B
- Native vision merged into one endpoint
- KV cache crushed: ~1/4 the HBM vs last one, 437× smaller than their first model
huggingface.co/deepseek-ai/...
- Asymmetric Causal-Encoder-Decoder: 550B MoE, input 8B / output 16B
- Native vision merged into one endpoint
- KV cache crushed: ~1/4 the HBM vs last one, 437× smaller than their first model
deepseek-ai/DeepSeek-V4.1-Flash · Hugging Face
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
huggingface.co
September 10, 2026 at 7:34 AM
DeepSeek v4.1 Flash is just another level 🤯
huggingface.co/deepseek-ai/...
- Asymmetric Causal-Encoder-Decoder: 550B MoE, input 8B / output 16B
- Native vision merged into one endpoint
- KV cache crushed: ~1/4 the HBM vs last one, 437× smaller than their first model
huggingface.co/deepseek-ai/...
- Asymmetric Causal-Encoder-Decoder: 550B MoE, input 8B / output 16B
- Native vision merged into one endpoint
- KV cache crushed: ~1/4 the HBM vs last one, 437× smaller than their first model
OCR for Japanese manga, Swedish handwriting or Arabic print?
There’s a growing range of OCR models on @hf.co: VLMs, dedicated text recognisers and complete OCR pipelines.
I’ve gathered 41 models into four collections, with short notes to help you choose:
huggingface.co/collections/...
There’s a growing range of OCR models on @hf.co: VLMs, dedicated text recognisers and complete OCR pipelines.
I’ve gathered 41 models into four collections, with short notes to help you choose:
huggingface.co/collections/...
OCR on the Hub - a davanstrien Collection
Curated OCR models for documents, languages, handwriting and text in images. Browse four collections with short practical notes.
huggingface.co
September 7, 2026 at 3:43 PM
OCR for Japanese manga, Swedish handwriting or Arabic print?
There’s a growing range of OCR models on @hf.co: VLMs, dedicated text recognisers and complete OCR pipelines.
I’ve gathered 41 models into four collections, with short notes to help you choose:
huggingface.co/collections/...
There’s a growing range of OCR models on @hf.co: VLMs, dedicated text recognisers and complete OCR pipelines.
I’ve gathered 41 models into four collections, with short notes to help you choose:
huggingface.co/collections/...
A 15.9M-parameter specialist OCR model outperforms many VLMs hundreds of times larger.
On a 2,165-page historical eval dataset, Kraken PP-OCRv6 ranks #4 on reading CER and #1 when long-s, ligatures and case are preserved.
huggingface.co/spaces/fineb...
On a 2,165-page historical eval dataset, Kraken PP-OCRv6 ranks #4 on reading CER and #1 when long-s, ligatures and case are preserved.
huggingface.co/spaces/fineb...
September 3, 2026 at 2:53 PM
A 15.9M-parameter specialist OCR model outperforms many VLMs hundreds of times larger.
On a 2,165-page historical eval dataset, Kraken PP-OCRv6 ranks #4 on reading CER and #1 when long-s, ligatures and case are preserved.
huggingface.co/spaces/fineb...
On a 2,165-page historical eval dataset, Kraken PP-OCRv6 ranks #4 on reading CER and #1 when long-s, ligatures and case are preserved.
huggingface.co/spaces/fineb...
Reposted by Daniel van Strien
📈 New blog post: Training and Finetuning Multi-Vector Embedding Models with Sentence Transformers.
As a practical example, I finetuned a ColBERT-style model for medical retrieval. 14.5 hours on one RTX 3090, and it beats every general-purpose retriever I could find.
Thread 🧵
As a practical example, I finetuned a ColBERT-style model for medical retrieval. 14.5 hours on one RTX 3090, and it beats every general-purpose retriever I could find.
Thread 🧵
August 26, 2026 at 1:56 PM
📈 New blog post: Training and Finetuning Multi-Vector Embedding Models with Sentence Transformers.
As a practical example, I finetuned a ColBERT-style model for medical retrieval. 14.5 hours on one RTX 3090, and it beats every general-purpose retriever I could find.
Thread 🧵
As a practical example, I finetuned a ColBERT-style model for medical retrieval. 14.5 hours on one RTX 3090, and it beats every general-purpose retriever I could find.
Thread 🧵
Every illustration in the Encyclopaedia Britannica dataset now has an instance mask: 411,385 cut-outs from 115,293 pages, 1768–1929, each linked to its full-resolution scan. Public domain, no image generation involved. Work in progress.
August 26, 2026 at 7:19 PM
Every illustration in the Encyclopaedia Britannica dataset now has an instance mask: 411,385 cut-outs from 115,293 pages, 1768–1929, each linked to its full-resolution scan. Public domain, no image generation involved. Work in progress.
Uploaded a dataset of 115,293 illustrated pages from the Encyclopaedia Britannica, 1st edition (1768) to 14th (1929) to the Hub
huggingface.co/datasets/big...
huggingface.co/datasets/big...
August 25, 2026 at 5:01 PM
Uploaded a dataset of 115,293 illustrated pages from the Encyclopaedia Britannica, 1st edition (1768) to 14th (1929) to the Hub
huggingface.co/datasets/big...
huggingface.co/datasets/big...
Every illustration in the British Library's 19th-century book images dataset now has an instance mask: 1.02M clean cutouts of public-domain engravings, no image generation involved!
One small segmentation model, $3.24 of compute. Model, masks and a search Space with cutout view all open.
One small segmentation model, $3.24 of compute. Model, masks and a search Space with cutout view all open.
August 20, 2026 at 10:42 AM
Every illustration in the British Library's 19th-century book images dataset now has an instance mask: 1.02M clean cutouts of public-domain engravings, no image generation involved!
One small segmentation model, $3.24 of compute. Model, masks and a search Space with cutout view all open.
One small segmentation model, $3.24 of compute. Model, masks and a search Space with cutout view all open.
Synthetic data at scale without owning a GPU: datatrove's new Jobs backend + Qwen3.8-27B → 35,837 length-controllable TL;DRs of Hugging Face cards, $0.43 per 1,000. Full guide: danielvanstrien.xyz/posts/2026/d...
Distilling Qwen3.8 with datatrove on Hugging Face Jobs – Daniel van Strien
datatrove’s new Jobs backend plus a days-old 27B teacher: regenerating a 35,837-summary training dataset in one afternoon for $15.58, with a calibration-first workflow and length-controllable outputs.
danielvanstrien.xyz
August 18, 2026 at 8:24 AM
Synthetic data at scale without owning a GPU: datatrove's new Jobs backend + Qwen3.8-27B → 35,837 length-controllable TL;DRs of Hugging Face cards, $0.43 per 1,000. Full guide: danielvanstrien.xyz/posts/2026/d...
FineBooks wants to help re-OCR the public domain to unlock a rich source of data for AI, for researchers, and for enjoyment.
Step one: work out which modern OCR models are actually good enough.
Step one: work out which modern OCR models are actually good enough.
August 10, 2026 at 12:18 PM
FineBooks wants to help re-OCR the public domain to unlock a rich source of data for AI, for researchers, and for enjoyment.
Step one: work out which modern OCR models are actually good enough.
Step one: work out which modern OCR models are actually good enough.
Reposted by Daniel van Strien
Are open OCR models good enough to unlock historical knowledge? @danielvanstrien.bsky.social and I built a leaderboard using six expert-transcribed volumes from the @biodivlibrary.bsky.social sky.social. Meet FineBooks, from @hf.co and @eleutherai.bsky.social! huggingface.co/blog/fineboo...
FineBooks: are open OCR models good enough to unlock historical knowledge?
A Blog post by FineBooks on Hugging Face
huggingface.co
August 10, 2026 at 12:03 PM
Are open OCR models good enough to unlock historical knowledge? @danielvanstrien.bsky.social and I built a leaderboard using six expert-transcribed volumes from the @biodivlibrary.bsky.social sky.social. Meet FineBooks, from @hf.co and @eleutherai.bsky.social! huggingface.co/blog/fineboo...
Uploaded a dataset of 1,080,814 public domain images, mostly from 19th-century books, to the Hugging Face Hub: huggingface.co/datasets/big...
You can also do semantic search against the images here: huggingface.co/spaces/davan...
You can also do semantic search against the images here: huggingface.co/spaces/davan...
August 7, 2026 at 5:18 PM
Uploaded a dataset of 1,080,814 public domain images, mostly from 19th-century books, to the Hugging Face Hub: huggingface.co/datasets/big...
You can also do semantic search against the images here: huggingface.co/spaces/davan...
You can also do semantic search against the images here: huggingface.co/spaces/davan...
Public data on which coding agents actually use the Hugging Face Hub, refreshed monthly.
Claude Code now sends a majority of everything the Hub can attribute to a named agent. In May, it led on 4 days out of 30. In July, all 30.
huggingface.co/datasets/hug...
Claude Code now sends a majority of everything the Hub can attribute to a named agent. In May, it led on 4 days out of 30. In July, all 30.
huggingface.co/datasets/hug...
August 3, 2026 at 11:48 AM
Public data on which coding agents actually use the Hugging Face Hub, refreshed monthly.
Claude Code now sends a majority of everything the Hub can attribute to a named agent. In May, it led on 4 days out of 30. In July, all 30.
huggingface.co/datasets/hug...
Claude Code now sends a majority of everything the Hub can attribute to a named agent. In May, it led on 4 days out of 30. In July, all 30.
huggingface.co/datasets/hug...
A film catalogue tells you what a film is about, not what happens inside it.
So: 1,864 public-domain Prelinger films, described every ~60 seconds by an open 2B video model. 23,148 searchable moments.
Search "typing on a computer keyboard", land on the second it happens.
So: 1,864 public-domain Prelinger films, described every ~60 seconds by an open 2B video model. 23,148 searchable moments.
Search "typing on a computer keyboard", land on the second it happens.
July 31, 2026 at 11:44 AM
A film catalogue tells you what a film is about, not what happens inside it.
So: 1,864 public-domain Prelinger films, described every ~60 seconds by an open 2B video model. 23,148 searchable moments.
Search "typing on a computer keyboard", land on the second it happens.
So: 1,864 public-domain Prelinger films, described every ~60 seconds by an open 2B video model. 23,148 searchable moments.
Search "typing on a computer keyboard", land on the second it happens.
New recipe: timestamped video captions on @hf.co Jobs.
Point it at a bucket of videos → parquet dataset out: scene descriptions + second-precise events.
~$0.05 per hour of footage on a single A10G (Marlin-2B on vLLM).
huggingface.co/datasets/uv-...
Point it at a bucket of videos → parquet dataset out: scene descriptions + second-precise
~$0.05 per hour of footage on a single A10G (Marlin-2B on vLLM).
huggingface.co/datasets/uv-...
July 30, 2026 at 8:31 AM
New recipe: timestamped video captions on @hf.co Jobs.
Point it at a bucket of videos → parquet dataset out: scene descriptions + second-precise events.
~$0.05 per hour of footage on a single A10G (Marlin-2B on vLLM).
huggingface.co/datasets/uv-...
Point it at a bucket of videos → parquet dataset out: scene descriptions + second-precise
~$0.05 per hour of footage on a single A10G (Marlin-2B on vLLM).
huggingface.co/datasets/uv-...
Reposted by Daniel van Strien
Small open models are getting genuinely good at document parsing: OvisOCR2 (0.9B, Apache 2.0) is claiming SOTA on OmniDocBench v1.6.
Day-1 recipe: OCR a whole HF image dataset — digitised newspapers, archives, zines — to markdown with one command.
huggingface.co/datasets/uv-scripts/ocr
Day-1 recipe: OCR a whole HF image dataset — digitised newspapers, archives, zines — to markdown with one command.
huggingface.co/datasets/uv-scripts/ocr
July 14, 2026 at 3:24 PM
Small open models are getting genuinely good at document parsing: OvisOCR2 (0.9B, Apache 2.0) is claiming SOTA on OmniDocBench v1.6.
Day-1 recipe: OCR a whole HF image dataset — digitised newspapers, archives, zines — to markdown with one command.
huggingface.co/datasets/uv-scripts/ocr
Day-1 recipe: OCR a whole HF image dataset — digitised newspapers, archives, zines — to markdown with one command.
huggingface.co/datasets/uv-scripts/ocr
Small open models are getting genuinely good at document parsing: OvisOCR2 (0.9B, Apache 2.0) is claiming SOTA on OmniDocBench v1.6.
Day-1 recipe: OCR a whole HF image dataset — digitised newspapers, archives, zines — to markdown with one command.
huggingface.co/datasets/uv-scripts/ocr
Day-1 recipe: OCR a whole HF image dataset — digitised newspapers, archives, zines — to markdown with one command.
huggingface.co/datasets/uv-scripts/ocr
July 14, 2026 at 3:24 PM
Small open models are getting genuinely good at document parsing: OvisOCR2 (0.9B, Apache 2.0) is claiming SOTA on OmniDocBench v1.6.
Day-1 recipe: OCR a whole HF image dataset — digitised newspapers, archives, zines — to markdown with one command.
huggingface.co/datasets/uv-scripts/ocr
Day-1 recipe: OCR a whole HF image dataset — digitised newspapers, archives, zines — to markdown with one command.
huggingface.co/datasets/uv-scripts/ocr
Reposted by Daniel van Strien
Reading the Archive by Machine:
An OCR Benchmark for Historians, 1612–1921
Here is version 1 of a working paper on the new OCR tools that are transforming digital history.
working-papers-in-critical-search.github.io/paper-004-oc...
An OCR Benchmark for Historians, 1612–1921
Here is version 1 of a working paper on the new OCR tools that are transforming digital history.
working-papers-in-critical-search.github.io/paper-004-oc...
Reading the Archive by Machine – Working Papers in Critical Search
A benchmark of six OCR systems (Tesseract, olmOCR 2, Chandra 2, Infinity Parser 2, GLM-OCR, and Gemini 3.5 Flash) on human-transcribed archival documents spanning 1612–1921: early-modern print, ninete...
working-papers-in-critical-search.github.io
July 10, 2026 at 4:21 PM
Reading the Archive by Machine:
An OCR Benchmark for Historians, 1612–1921
Here is version 1 of a working paper on the new OCR tools that are transforming digital history.
working-papers-in-critical-search.github.io/paper-004-oc...
An OCR Benchmark for Historians, 1612–1921
Here is version 1 of a working paper on the new OCR tools that are transforming digital history.
working-papers-in-critical-search.github.io/paper-004-oc...
Reposted by Daniel van Strien
I don't have it yet, but there's an econ history paper in these transcripts or transcripts like them (that we can now turn into textual data at the cost of pennies per hour)
Open ASR models with speaker diarization are now fast and cheap: I diarized 174 hours of Apollo 11 mission audio (the real July 1969 NASA tapes) for $9.46 with a 0.9B open model. Search it, hear any moment on the original tape:
huggingface.co/spaces/davan...
huggingface.co/spaces/davan...
July 9, 2026 at 6:32 PM
I don't have it yet, but there's an econ history paper in these transcripts or transcripts like them (that we can now turn into textual data at the cost of pennies per hour)
Open ASR models with speaker diarization are now fast and cheap: I diarized 174 hours of Apollo 11 mission audio (the real July 1969 NASA tapes) for $9.46 with a 0.9B open model. Search it, hear any moment on the original tape:
huggingface.co/spaces/davan...
huggingface.co/spaces/davan...
July 9, 2026 at 4:41 PM
Open ASR models with speaker diarization are now fast and cheap: I diarized 174 hours of Apollo 11 mission audio (the real July 1969 NASA tapes) for $9.46 with a 0.9B open model. Search it, hear any moment on the original tape:
huggingface.co/spaces/davan...
huggingface.co/spaces/davan...
I ran 10 newer OCR models on @ai2.bsky.social's olmOCR-bench "old scans" subset.
The ranking flips depending on what you actually want.
The ranking flips depending on what you actually want.
July 7, 2026 at 4:36 PM
I ran 10 newer OCR models on @ai2.bsky.social's olmOCR-bench "old scans" subset.
The ranking flips depending on what you actually want.
The ranking flips depending on what you actually want.
Coding agents are real users of the @hf.co Hub!
They're searching for models, building and pushing datasets, training models on Jobs, spinning up Spaces...
Now there's public data: each agent's share of Hub traffic, updated monthly 👇
They're searching for models, building and pushing datasets, training models on Jobs, spinning up Spaces...
Now there's public data: each agent's share of Hub traffic, updated monthly 👇
July 2, 2026 at 3:37 PM
Coding agents are real users of the @hf.co Hub!
They're searching for models, building and pushing datasets, training models on Jobs, spinning up Spaces...
Now there's public data: each agent's share of Hub traffic, updated monthly 👇
They're searching for models, building and pushing datasets, training models on Jobs, spinning up Spaces...
Now there's public data: each agent's share of Hub traffic, updated monthly 👇