Annotix: An Integrated Desktop Platform for Multi-Modal Data Annotation, Collaborative Labeling, and End-to-End Machine Learning Training

Annotated dataset preparation remains a critical bottleneck in machine learning (ML) pipelines. Existing annotation tools—cloud-hosted services, self-hosted web applications, and lightweight desktop editors—each cover part of this workflow, but few combine broad annotation-type support with offline operation, integrated training, and serverless collaboration. We present Annotix, an open-source, cross-platform desktop application that brings the complete ML data preparation workflow—annotation, training, model-assisted labeling, and review—into a single privacy-preserving environment. It is implemented with a Rust/Tauri 2 backend and a React 19 frontend, and runs locally without server infrastructure or cloud services. Annotix was evaluated through three complementary studies. First, a comparative feature analysis positioned the platform against established annotation tools. Second, a controlled efficiency experiment compared Annotix, CVAT, and Label Studio, in which three evaluators annotated 60 synthetic images across bounding box and mask tasks, analyzed with Kruskal–Wallis and Dunn–Bonferroni post hoc tests. Third, a heuristic usability evaluation assessed standardized tasks on real medical images (retinal fundus and otoscopic images). Results indicate that Annotix achieves annotation efficiency competitive with established tools while offering broader integrated coverage, including end-to-end model training and serverless peer-to-peer collaboration. Annotix is freely available under the MIT license and targets privacy-sensitive domains such as medical imaging and ecological monitoring.