---
title: "2026-08-24 PESO dataset research"
description: "Researched the public PESO histopathology dataset's packaging, licensing, checksums, and remote-access characteristics."
image: "https://docs.virdx.dev/img/virdx-social-card.png"
---

> Documentation Index
> Fetch the complete documentation index at: https://docs.virdx.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# 2026-08-24 PESO dataset research

## Work Done

- Inspected Zenodo record 5137717 and its API metadata for the PESO prostate epithelium segmentation dataset.
- Catalogued all 18 deposit files, exact byte sizes, MD5 checksums, direct download endpoints, formats, stated image resolutions, and dataset semantics.
- Used byte-range reads of ZIP central directories to verify archive member counts and filenames without downloading the approximately 138 GB deposit.
- Assessed one-file-at-a-time access: Zenodo serves archive-level objects with HTTP byte ranges, but does not expose inner ZIP members as standalone resources.

## Lessons Learned: Pitfalls

- Two archives named `peso_testset_png*.zip` are described as PNG exports, but their members are actually `.jpg`; relying only on the record prose would miss this discrepancy.
- Standard download tools treat each 12–15 GB WSI archive as one object. Reading one member remotely requires a range-aware ZIP implementation and careful handling of ZIP64 and deflated members.

## Lessons Learned: Improvements

- A reusable remote ZIP inventory tool supporting ZIP64 would speed future large public-dataset audits while avoiding full downloads.
- Dataset ingestion guidance should distinguish server-side per-file URLs from client-side range extraction of members inside compressed archives.

Source: https://docs.virdx.dev/knowledge/inbox/2026-08-24-histo-peso-dataset-research/index.mdx