---
license: cc-by-nc-sa-4.0
library_name: pytorch
tags:
  - histology
  - neuroscience
  - hippocampus
  - bigbrain
  - cell-segmentation
  - density-estimation
  - non-commercial
---

# CALHippo Model Weights

<div align="center">
  <img src="CALHippo_logo_transparent.png" alt="CALHippo Logo" width="400" />
</div>

This repository contains released model artifacts for the [CALHippo Framework](https://github.com/AImageLab-zip/CALHippo-Framework),
including segmentation, density-estimation, and classification components used
by the public reproducibility path.

## License and Terms of Use

The CALHippo model weights in this repository are released under
Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International
(CC BY-NC-SA 4.0).

These weights were trained and/or fine-tuned using BigBrain-derived
histological data. They are intended for non-commercial academic research use
only. Redistribution, adaptation, and sharing of derived versions must comply
with CC BY-NC-SA 4.0 and with any applicable upstream dataset, model, and
software licenses.

This repository does not redistribute UNI2-h weights. UNI2-h is a gated upstream
model distributed by MahmoodLab under its own access terms and license. Users
who want to run the UNI2-h classification path must request access from the
upstream provider and authenticate locally.

## Included Artifacts

```text
density_estimation/
  short_unet/
    final_density_model.pth
    *.yaml

segmentation/
  cellpose/
    finetune_v4_astrocytes_big_brain
  hovernet/
    net_epoch=20.tar
  instanseg/
    instanseg.pt
  stardist/
    config.json
    thresholds.json
    weights_best.h5

classification/
  ml_classifier/
    model.joblib
    metadata.json
    metrics.json
```
## Not Included
`classification/feature_encoder/uni2h/`

UNI2-h is not redistributed. The CALHippo setup script downloads or loads UNI2-h
only after the user has obtained upstream access and authenticated with Hugging
Face.

## Attribution

If you use these model weights, please cite the CALHippo framework and the
upstream resources used by the pipeline, including BigBrain and the relevant
model families:

BigBrain
Selected 1 micron BigBrain histological sections
Cellpose / Cellpose-SAM
HoVer-Net
InstanSeg
StarDist
UNI / UNI2-h, when using the UNI2-h classification path

Please also retain attribution and license notices for upstream software and
model components used by the CALHippo framework.

## Intended Use

These weights are intended for non-commercial academic research on histological
image analysis, hippocampal cellular annotation, segmentation, classification,
density estimation, and mesoscale reconstruction.

They are not intended for clinical diagnosis, medical decision-making, or
commercial deployment.

## Loading with the CALHippo framework

The recommended way to obtain these artifacts is through the [CALHippo Framework](https://github.com/AImageLab-zip/CALHippo-Framework)