Latent Bridge Matching (LBM) is a new, versatile and scalable method proposed in LBM: Latent Bridge Matching for Fast Image-to-Image Translation that relies on Bridge Matching in a latent space to achieve fast image-to-image translation. This model was trained to estimate the depth map from a given input image. See also our live demo for image relighting and official Github repo.

To use this model you need first to install the associated lbm library by running the following

pip install git+https://github.com/gojasper/LBM.git

Then, you can infer with the model on your input images

import torch
from PIL import Image
from lbm.inference import evaluate, get_model

# Load model
model = get_model(
  "jasperai/LBM_depth",
  torch_dtype=torch.bfloat16,
  device="cuda",
)

# Load a source image
source_image = Image.open("your_image")

# Perform inference
output_image = evaluate(model, source_image, num_sampling_steps=1)

This model achieves the following metrics

Metrics	AbsRel ↓	δ1 ↑	δ2 ↑
NYUv2	5.6	97.2	99.2
Kitti	9.4	93.0	98.9
ETH3D	6.3	96.5	99.3
ScanNet	5.7	97.0	99.2
DIODE	30.3	77.5	89.3

This code is released under the Creative Commons BY-NC 4.0 license.

If you find this work useful or use it in your research, please consider citing us

@article{chadebec2025lbm,
      title={LBM: Latent Bridge Matching for Fast Image-to-Image Translation}, 
      author={Clément Chadebec and Onur Tasar and Sanjeev Sreetharan and Benjamin Aubin},
      year={2025},
      journal = {arXiv preprint arXiv:2503.07535},
}