Companies / Regulator
Cefas
The UK government's Centre for Environment, Fisheries and Aquaculture Science, which buys image annotation. Recorded here: 3 public prices.
cefas.co.ukLatest record 10 Mar 2025
Public prices involving Cefas
- Marine-litter annotation of drone survey images for partner countries, to train a litter-detection model, plus training for in-country partners (Cefas, UK government, OCPP programme)
The third Cefas annotation buy in the series (after CEFAS21-99 and CEFAS23-100); it also pays for training partners to run annotation themselves.
The payer's own page, UK Contracts Finder (official API record), Cefas award notice CEFAS24-111, marine litter image annotation services
£47,933per contract (Dec 2024 to Mar 2025)
- About 7,000 drone images of Ghanaian beaches annotated for 89 marine-litter classes to train a detection model (Cefas, UK government, OCPP programme)
A UK government science agency buying training data for its own model from a Ghana-based annotation firm; the price also covers three inter-rater assessments, so the per-image figure is all-in.
The payer's own page, UK Contracts Finder (official API record), Cefas award notice CEFAS23-100, Ghanaian image annotation of marine litter images
£46,500per contract (about 7,000 images, Sep 2023 to Mar 2024)
- Pixelwise segmentation of at least 5,000 high-resolution drone images of UK beaches into at least 50 marine-litter classes, for a machine-learning model (Cefas, UK government)
The notice sets a minimum scope, so the per-image figure is a ceiling; the supplier, HITL EOOD, is a Bulgarian human-in-the-loop annotation firm.
The payer's own page, UK Contracts Finder (official API record), Cefas award notice CEFAS21-99, OCPP image annotation for machine learning algorithm
£30,000per contract (at least 5,000 images, Dec 2021 to Mar 2022)
Cite this
AIData.Broker, “Cefas in the AI training data market”, reviewed 10 Mar 2025. https://www.aidata.markets/companies/cefas
Free to reuse under CC BY 4.0 if you credit AIData.Broker and link to this page. Each row names the publication that stated it, and words quoted from a source keep that source's own terms.