arXiv subjects

Colin Groth

Publications and source records attributed to Colin Groth.

3 recordsLinked to original sources

arXiv2026

Infinite Gaze Generation for Videos with Autoregressive Diffusion

Predicting human gaze in video is fundamental to advancing scene understanding and multimodal interaction. While traditional saliency maps provide spatial probability distributions and scanpaths offer ordered fixations, both abstractions often collapse the fine-grained temporal dynamics of raw gaze. Furthermore, existing models are typically constrained to short-term windows ($\approx$ 3-5s), failing to capture the long-range behavioral dependencies inherent in real-world content. We present a generative framework for infinite-horizon raw gaze prediction in videos of arbitrary length. By leveraging an autoregressive diffusion model, we synthesize gaze trajectories characterized by continuous spatial coordinates and high-resolution timestamps. Our model is conditioned on a saliency-aware visual latent space. Quantitative and qualitative evaluations demonstrate that our approach significantly outperforms existing approaches in long-range spatio-temporal accuracy and trajectory realism.

cs.CV↗

arXiv2025

MILO: A Lightweight Perceptual Quality Metric for Image and Latent-Space Optimization

We present MILO (Metric for Image- and Latent-space Optimization), a lightweight, multiscale, perceptual metric for full-reference image quality assessment (FR-IQA). MILO is trained using pseudo-MOS (Mean Opinion Score) supervision, in which reproducible distortions are applied to diverse images and scored via an ensemble of recent quality metrics that account for visual masking effects. This approach enables accurate learning without requiring large-scale human-labeled datasets. Despite its compact architecture, MILO outperforms existing metrics across standard FR-IQA benchmarks and offers fast inference suitable for real-time applications. Beyond quality prediction, we demonstrate the utility of MILO as a perceptual loss in both image and latent domains. In particular, we show that spatial masking modeled by MILO, when applied to latent representations from a VAE encoder within Stable Diffusion, enables efficient and perceptually aligned optimization. By combining spatial masking with a curriculum learning strategy, we first process perceptually less relevant regions before progressively shifting the optimization to more visually distorted areas. This strategy leads to significantly improved performance in tasks like denoising, super-resolution, and face restoration, while also reducing computational overhead. MILO thus functions as both a state-of-the-art image quality metric and as a practical tool for perceptual optimization in generative pipelines.

cs.CV↗

arXiv2022

Wavelet-Based Fast Decoding of 360-Degree Videos

In this paper, we propose a wavelet-based video codec specifically designed for VR displays that enables real-time playback of high-resolution 360{\deg} videos. Our codec exploits the fact that only a fraction of the full 360{\deg} video frame is visible on the display at any time. To load and decode the video viewport-dependently in real time, we make use of the wavelet transform for intra- as well as inter-frame coding. Thereby, the relevant content is directly streamed from the drive, without the need to hold the entire frames in memory. With an average of 193 frames per second at 8192x8192-pixel full-frame resolution, the conducted evaluation demonstrates that our codec's decoding performance is up to 272% higher than that of the state-of-the-art video codecs H.265 and AV1 for typical VR displays. By means of a perceptual study, we further illustrate the necessity of high frame rates for a better VR experience. Finally, we demonstrate how our wavelet-based codec can also directly be used in conjunction with foveation for further performance increase.

cs.GR↗