Lightweight multimodal classifier for scientific poster quality control in open repositories.
Part of the quality control pipeline for posters.science, a platform for making scientific conference posters Findable, Accessible, Interoperable, and Reusable (FAIR).
Developed by the FAIR Data Innovations Hub at the California Medical Innovations Institute (CalMI2).
Open repositories like Zenodo and Figshare host tens of thousands of records labeled as scientific posters. However, approximately 20% of these records are mislabeled — containing multi-page papers, conference proceedings, abstract booklets, slide decks, or other non-poster documents. This label noise is a significant barrier to automated poster processing at scale.
PosterSentry classifies PDFs using three complementary feature channels concatenated into a 542-dimensional vector:
| Channel | Features | Dimensions | Signal |
|---|---|---|---|
| Text | model2vec (potion-base-32M) embedding | 512 | Semantic content |
| Visual | Color stats, edge density, FFT spatial complexity, whitespace | 15 | Visual layout |
| Structural | Page count, area, font diversity, text blocks, density | 15 | PDF geometry |
A StandardScaler normalizes all features (preventing the 512-d text embedding from drowning out structural/visual signal), then a LogisticRegression classifier produces the final prediction.
The classifier head is a single linear layer stored as a numpy .npz file (10 KB). Inference is pure numpy — no GPU or deep learning framework required.
Validated on 3,606 real scientific documents (zero synthetic data):
| Metric | Value |
|---|---|
| Accuracy | 87.3% |
| F1 (poster) | 87.1% |
| F1 (non-poster) | 87.4% |
| Precision (poster) | 88.2% |
| Recall (poster) | 85.9% |
| Inference speed | < 1 sec/PDF (CPU) |
Applied to 30,205 PDFs from Zenodo and Figshare, PosterSentry classified 80.2% as true posters and 19.8% as non-posters, with mean confidence of 0.799.
| Feature | Coefficient | Signal |
|---|---|---|
size_per_page_kb |
+7.65 | Posters are dense, high-res single pages |
page_count |
-5.49 | More pages = not a poster |
file_size_kb |
-5.44 | Multi-page docs are bigger overall |
is_landscape |
+0.98 | Some posters are landscape |
color_diversity |
+0.95 | Posters are visually rich |
edge_density |
+0.79 | More visual edges in posters |
pip install poster-sentry# Classify a single PDF
poster-sentry classify document.pdf
# Classify multiple PDFs
poster-sentry classify *.pdf --output results.tsv
# Print model info
poster-sentry infofrom poster_sentry import PosterSentry
sentry = PosterSentry()
sentry.initialize()
# Classify a PDF (uses text + visual + structural features)
result = sentry.classify("document.pdf")
print(f"Is poster: {result['is_poster']}, Confidence: {result['confidence']:.2f}")
# Batch classification
results = sentry.classify_batch(["poster1.pdf", "paper.pdf", "newsletter.pdf"])
# Text-only classification (no PDF needed)
result = sentry.classify_text("Title: My Poster\nAuthors: ...")PosterSentry sits at the front of the posters.science pipeline — it screens incoming PDFs before expensive LLM-based extraction:
PDF Input
|
v
PosterSentry --> poster2json --> FAIR output
(classify: poster?) (Llama 3.1 8B structured extraction) (poster-json-schema)
| Requirement | Value |
|---|---|
| CPU | Any modern CPU (no GPU needed) |
| RAM | 4 GB+ |
| Python | 3.10+ |
| Model size | 10 KB head + ~60 MB embeddings (downloaded once) |
| Resource | Description |
|---|---|
| poster-sentry (HuggingFace) | Model weights and config |
| poster-sentry-training-data (HuggingFace) | Training dataset (3,606 samples) |
| poster-sentry-training (GitHub) | Training code and replication |
| poster2json | Poster to structured JSON extraction |
| posters.science | Platform |
git clone https://github.com/fairdataihub/poster-sentry.git
cd poster-sentry
pip install -e ".[dev]"
pytest@software{poster_sentry_2026,
title = {PosterSentry: Multimodal Scientific Poster Classifier},
author = {O'Neill, James and Soundarajan, Sanjay and Portillo, Dorian and Patel, Bhavesh},
year = {2026},
url = {https://github.com/fairdataihub/poster-sentry},
note = {Part of the posters.science initiative at FAIR Data Innovations Hub}
}MIT License. See LICENSE for details.
- FAIR Data Innovations Hub at California Medical Innovations Institute (CalMI2)
- posters.science platform
- MinishLab for the model2vec embedding backbone
- Funded by The Navigation Fund — "Poster Sharing and Discovery Made Easy"