{"slug": "gazedit-gaze-accurate-diffusion-image-generation-for-eye-tracking-via-spatial", "title": "GazeDiT: Gaze-Accurate Diffusion Image Generation for Eye Tracking via Spatial Conditioning", "summary": "GazeDiT, a diffusion model that generates eye-tracking images for a requested 4D binocular gaze via an internally constructed spatial condition, achieves substantially lower tail gaze-label error than other diffusion baselines, approaching the error of the same frozen gaze estimator on real images. The model uses a frozen SegFormer to extract pupil and iris geometry during training and a physical eye renderer at inference to sample gaze-consistent geometries without a source image. Its generated data improved a downstream eye tracker, reducing gaze error on difficult cases from 3.05 degrees to 2.80 degrees in the smallest cohort.", "body_md": "arXiv:2609.17814v1 Announce Type: new \nAbstract: Diffusion models are increasingly used to generate synthetic training data, but precise label control remains difficult when the conditioning signal is low-dimensional and coarse. Text-conditioned images are judged by broad prompt consistency, whereas supervised training requires precise correspondence between each image and its numerical label. This is challenging in eye tracking, where a 4D binocular gaze is expressed through subtle, spatially localized pupil and iris geometry. We introduce GazeDiT, a diffusion model that generates images for a requested 4D gaze through an internally constructed spatial condition that grounds the global gaze label in this local geometry. During training, a frozen SegFormer extracts pupil/iris geometry from diverse real images, allowing the model to learn realistic appearance conditioned on that geometry. At inference, a physical eye renderer samples gaze-consistent geometries by varying anatomy and camera state, enabling diverse synthesis without a source image. GazeDiT achieves substantially lower tail gaze-label error than other diffusion baselines, approaching the error of the same frozen gaze estimator on real images. Its generated data also improves the downstream eye tracker, reducing gaze error on difficult cases from 3.05{\\deg} to 2.80{\\deg} in the smallest cohort.", "url": "https://wpnews.pro/news/gazedit-gaze-accurate-diffusion-image-generation-for-eye-tracking-via-spatial", "canonical_source": "https://arxiv.org/abs/2609.17814", "published_at": "2026-09-17 04:00:00+00:00", "updated_at": "2026-09-17 04:26:43.874422+00:00", "lang": "en", "topics": ["generative-ai", "computer-vision", "ai-research", "machine-learning"], "entities": ["GazeDiT", "SegFormer", "arXiv"], "alternates": {"html": "https://wpnews.pro/news/gazedit-gaze-accurate-diffusion-image-generation-for-eye-tracking-via-spatial", "markdown": "https://wpnews.pro/news/gazedit-gaze-accurate-diffusion-image-generation-for-eye-tracking-via-spatial.md", "text": "https://wpnews.pro/news/gazedit-gaze-accurate-diffusion-image-generation-for-eye-tracking-via-spatial.txt", "jsonld": "https://wpnews.pro/news/gazedit-gaze-accurate-diffusion-image-generation-for-eye-tracking-via-spatial.jsonld"}}