https://chenbh.com/tags/computer-vision/Computer VisionHugo Blox Builder (https://hugoblox.com)en-usTue, 07 Jul 2026 00:00:00 +0000https://chenbh.com/media/icon_hu5000416120042106492.pnghttps://chenbh.com/tags/computer-vision/https://chenbh.com/publication/mammadov-variational-2026/Tue, 07 Jul 2026 00:00:00 +0000https://chenbh.com/publication/mammadov-variational-2026/<p><strong>Variational Flow Maps (VFM)</strong> recasts conditional generation as a problem of learning the right initial noise distribution for a pretrained or jointly trained one-step flow map. An observation-dependent adapter transforms simple noise before the flow map sends it to data space, enforcing the measurement while retaining the learned data prior.</p> <p>A variational training objective aligns the adapted noise with the conditional target and supports both amortized inverse-problem solving and reward-based alignment. Because the conditioning mechanism acts before generation, VFM avoids tracing and guiding the long sampling trajectories required by diffusion and iterative flow models.</p> <p>Across image inverse problems and ImageNet conditional-generation tasks, VFM produces diverse, well-calibrated samples in one or a few network evaluations. The results combine competitive fidelity with orders-of-magnitude faster sampling than iterative diffusion and flow baselines.</p>