Clément Chadebec @clementchadebec.bsky.social · 13/05/2025We also release ckpts for normal and depth estimation where the model should translate an input image into a normal or depth map. 🤗 Ckpt: huggingface.co/jasperai/LBM... 🤗 Ckpt: huggingface.co/jasperai/LBM... 010
Clément Chadebec @clementchadebec.bsky.social · 13/05/2025Today, we release on @hf.co, the model for object relighting where the model should relight a foreground object so that it blends perfectly with a target background. 🤗Ckpt: huggingface.co/jasperai/LBM... 🤗Demo: huggingface.co/spaces/jaspe... 110
Clément Chadebec @clementchadebec.bsky.social · 13/05/2025In the proposed LBM method, we propose to encode the source and target image into a latent space and then build a stochastic path called a Brownian Bridge between them. In particular, the stochasticity of these paths makes the method differ from flow matching. 110
Clément Chadebec @clementchadebec.bsky.social · 13/05/2025Latent Bridge Matching (LBM) is a method that aims at transporting a source distribution of images (e.g. pasted images) to a target distribution (e.g. relighted images). 🤗Demo: huggingface.co/spaces/jaspe... 110
Clément Chadebec @clementchadebec.bsky.social · 13/05/2025📢 Latent Bridge Matching update! 📢 New Jasper Research release 🚀 - 3 LBM models on @hf.co (relighting, depth, normals) - LBM training code Also see our @gradio-hf.bsky.social space for object relighting! 🤗 huggingface.co/jasperai/LBM... 💻 github.com/gojasper/LBM 📗 arxiv.org/abs/2503.07535 263
Clément Chadebec @clementchadebec.bsky.social · 13/03/2025… finally, we also consider common tasks such as normal and depth estimation where the model should translate an input image into a normal or depth map 130
Clément Chadebec @clementchadebec.bsky.social · 13/03/2025... we also derive a conditional framework of LBM and demonstrate its effectiveness by tackling the tasks of controllable image relighting and shadow generation 🕹️ … 120
Clément Chadebec @clementchadebec.bsky.social · 13/03/2025... and image restoration 🧹 where the model must transport the distribution of the degraded images to the distribution of the clean images … 120
Clément Chadebec @clementchadebec.bsky.social · 13/03/2025… but also object removal ✂️ where the model is trained to find a transport map from the masked images to the images without the objects … 120
Clément Chadebec @clementchadebec.bsky.social · 13/03/2025During training, we also introduce a pixel loss that consists of decoding the estimated target latent and comparing it to the real target image. We found that LPIPS works well in practice and speeds up domain shift. We validate the method for object relighting 🔦... 130
Clément Chadebec @clementchadebec.bsky.social · 13/03/2025We encode paired images into the latent space and bridge the latents 🌉. A timestep is drawn from a well chosen distribution. The latent on the trajectory at this timestep is then passed to the denoiser which predicts the drift of the associated Stochastic Differential Equation. 130
Clément Chadebec @clementchadebec.bsky.social · 13/03/2025In the proposed LBM method, we propose to encode the source and target image into a latent space and then build a stochastic path called a Brownian Bridge 🌉 between them. In particular, the stochasticity of these paths makes the method differ from flow matching. 140
Clément Chadebec @clementchadebec.bsky.social · 13/03/2025Excited to share our new research at Jasper Research! 🚀 LBM: Latent Bridge Matching for Fast Image-to-Image Translation Try out our @hf.co space for object relighting! 🤗 @gradio-hf.bsky.social demo: huggingface.co/spaces/jaspe... 👉 Paper: arxiv.org/abs/2503.07535 💻 Repo: github.com/gojasper/LBM 2207
Clément Chadebec @clementchadebec.bsky.social · 12/12/2024🎉 Very happy to share that Jasper Research's very first research paper, Flash Diffusion just got accepted to AAAI 2025. This paper introduces a novel and flexible distillation method designed to speed up image generation with diffusion models. 👉 Paper: arxiv.org/abs/2406.02347 283
