This use case develops methods and datasets for cattle welfare assessment using computer vision. Introduction The aim of the project is to develop an operational national Animal Welfare Verification System for assessing, verifying, and communicating animal welfare. The system shall bring added value to animal-based products which meet high animal welfare criteria and respond to the demands of market and consumers. The system is based on scientifically validated on-farm assessment protocols for cattle and third-party auditions. It will generate new business opportunities and enable expanding Finnish animal product trade to new customer segments and regions while strengthening its position in the current markets. Methodologies Advanced computer vision- based system for monitoring dairy cow welfare will be developed. The system will be based on data collected from several farms and annotated for animal masks, keypoints and tracks. AI assisted annotation workflows are being developed as part of the project. This use case extends the national work on cattle tracking using computer vision by creating guidelines for the use of developed advanced annotation workflows, datasets and AI methods for cow tracking. Output The project will produce 1) Computer vision models for semi-supervised cow tracking and derive welfare indicators from the data. 2) High-quality annotated dataset of cattle tracks for training of computer vision models combined with a public benchmark 3) publish an application and workflow for foundation model-based data annotation 4) develop semantic meta-model for sharing welfare data in a data space. Impact The use case results will be applicable to European dairy farms with indoor housing systems. Computer vision is currently seen as the key method for developing more accurate welfare monitoring and this use case will deliver open dataset and methodological progress. These results will be presented at multiple European events and the UC will seek to collaborate with relevant European research groups and projects. What I can offer Annotation workflow for animal tracking, workflows for model training, evaluation and use. What I need Collaborations on method development and shared datasets.

Source: Annual AgData Use Case Summit 2026 booklet.