TDWG2026
Home / Talks / SAL C
talk · Tuesday 22 September · 14:15–⁠14:30 · SAL C

Pinned insect digitization conveyor: image and data transcription workflows

Jessica Bird, Sylvia Orli · Digitizing Collections & Legacy Data

Recording time 5:48:07–6:02:02Open on Vimeo ↗

Summaries and key moments are machine-produced. Check the recording for the speaker’s exact words.

The short versionAn industrial conveyor workflow with QR-encoded metadata and automated skeletal records makes mass digitisation of pinned insects feasible.

Overview

What this was about

Jessica Bird described the Smithsonian NMNH Entomology conveyor project digitising pinned pollinators (bees, beetles, butterflies, moths, flies), funded for an estimated 360,000 specimens, in a collection of about 35 million specimens that would take 20 people 67 years to digitise traditionally. Pre-curation vets taxonomy in EMu and prints unit-tray header labels with QR-embedded taxonomy and identifiers; on the conveyor, labels are staged, specimens barcoded and three focus-stacked views plus label images captured, with metadata in JSON. Images pass QC in the open-source OSPRY tool, enter the DAMS, generate skeletal EMu records published to GBIF via IPT, while label images feed a separate contracted human transcription workflow.

Why it matters. Demonstrates throughput, cost and QC practices for mass digitisation of one of the hardest specimen types, relevant to anyone planning large entomology projects.

Key ideas

In the room

  • USNM Entomology is the second-largest entomology collection (about 35 million specimens, over 21 million pinned).
  • Project started 2025 on pinned pollinators, funded for about 360,000 specimens.
  • Header labels with QR-embedded taxonomy and unit tray identifiers streamline later record creation; drawer IDs name image folders.
  • Silhouette determines imaging sphere/resolution; three focus-stacked views (about 30 slices each) and a composite label image are taken.
  • QC in OSPRY samples 10-40% of images per folder with standardised error phrasing.
  • Skeletal records are created automatically in EMu and published to the collections website and GBIF.
  • Label orientation could not be automatically corrected, so label transcription was split into its own contracted human workflow.
Jump in

Notable moments

In their words

Transcript

Automatically generated captions can contain mistakes, especially in names and technical terms. Times are relative to the room recording.

Read the transcript ↓
Loading transcript…
Keep exploring

Related talks.

See the map ↗
talk · 25 Sept · 09:24–⁠09:36 · SAL A

Automated data transcription and Darwin Core field itemization from insect specimen labels using machine learning

Labellum goes beyond label OCR to itemise insect-label text into Darwin Core fields, with high accuracy for most fields but problems with catalogue numbers.

Torsten Dikow
talk · 22 Sept · 16:35–⁠16:40 · SAL B

From Collection Drawers to AI - A One-Week Recipe

A useful AI-ready digitisation setup can be built in a week from recycled materials and borrowed cameras.

Arianna Salili-James
talk · 24 Sept · 10:17–⁠10:30 · SAL C

Updates on Data Mobilization in the KU Entomology Collection

Moving to Specify 7 with a nightly Darwin Core archive feed made KU Entomology's data and images continuously available to GBIF, Symbiota portals and students.

Samanta Orellana
Programme abstract

Pinned Insect Digitization Conveyor: image and data transcription workflows

Jessica Bird, Sylvia Orli

The Smithsonian National Museum of Natural History (NMNH) is currently digitizing pollinating insects using a Pinned Insect Digitization Conveyor developed by Picturae. The digital workflow, developed by the Smithsonian Office of Digital and Innovation, the NMNH Informatics and Data Science Center, and the Department of Entomology, are reviewed here. Insects and labels are placed on the conveyor and up to 1,500 specimens are processed daily. Four images are captured of each insect - one of labels and three focus-stacked images of the specimen. Conveyor images are run through a technical review process and are screened by Picturae before delivery to Smithsonian. Files are organized into a folder system which hold all specimen images from an insect drawer. A second technical check, using the Smithsonian system, Osprey, is followed by review of random images by an individual. Imaging errors may include issues related to automated focus stacking and specimens being captured at the wrong resolution for their size. When a drawer has passed QC, images are delivered to the institutional DAMS. A script then picks up the images for inclusion into the NMNH database of record, EMu. Stub records are created, images are attached, and the record is then published. Label images are submitted to a different vendor for data transcription. Transcribed data are run through Osprey where a review of randomly selected data is checked for accuracy. When a drawer has passed QC, data is transformed for bulk import back into EMu. Data is appended to the original stub records and images are made available for Open Access (CC0), if applicable. The current project goal is the digitization of 365,000 specimens of pinned insects of which the majority are pollinating insects.

Submitted to the conference programme before the talk, so the talk itself may differ.