In this paper, we propose an efficient, fast, and versatile distillation method to accelerate the generation of pre-trained diffusion models. The method reaches state-of-the-art performances in terms of FID and CLIP-Score for few steps image generation on the COCO2014 and COCO2017 datasets, while requiring only several GPU hours of training and fewer trainable parameters than existing methods. In addition to its efficiency, the versatility of the method is also exposed across several tasks such as *text-to-image*, *inpainting*, *face-swapping*, *super-resolution* and using different backbones such as UNet-based denoisers (SD1.5, SDXL), DiT (Pixart) and MMDiT (SD3), as well as adapters. In all cases, the method allowed to reduce drastically the number of sampling steps while maintaining very high-quality image generation.

In this paper, we introduce Latent Bridge Matching (LBM), a new, versatile and scalable method that relies on Bridge Matching in a latent space to achieve fast image-to-image translation. We show that the method can reach state-of-the-art results for various image-to-image tasks using only a single inference step. In addition to its efficiency, we also demonstrate the versatility of the method across different image translation tasks such as object removal, normal and depth estimation, and object relighting. We also derive a conditional framework of LBM and demonstrate its effectiveness by tackling the tasks of controllable image relighting and shadow generation. We provide an implementation at https://github.com/gojasper/LBM.

The domain adaptation of satellite images has recently gained increasing attention to overcome the limited generalization abilities of machine learning models when segmenting large-scale satellite images. Most of the existing approaches seek for adapting the model from one domain to another. However, such single-source and single-target setting prevents the methods from being scalable solutions since, nowadays, multiple sources and target domains having different data distributions are usually available. Besides, the continuous proliferation of satellite images necessitates the classifiers to adapt to continuously increasing data. We propose a novel approach, coined DAugNet, for unsupervised, multisource, multitarget, and life-long domain adaptation of satellite images. It consists of a classifier and a data augmentor. The data augmentor, which is a shallow network, is able to perform style transfer between multiple satellite images in an unsupervised manner, even when new data are added over time. In each training iteration, it provides the classifier with diversified data, which makes the classifier robust to large data distribution difference between the domains. Our extensive experiments prove that DAugNet significantly better generalizes to new geographic locations than the existing approaches.

Although convolutional neural networks have been proven to be an effective tool to generate high quality maps from remote sensing images, their performance significantly deteriorates when there exists a large domain shift between training and test data. To address this issue, we propose a new data augmentation approach that transfers the style of test data to training data using generative adversarial networks. Our semantic segmentation framework consists in first training a U-net from the real training data and then fine-tuning it on the test stylized fake training data generated by the proposed approach. Our experimental results prove that our framework outperforms the existing domain adaptation methods.

Due to the various reasons, such as atmospheric effects and differences in acquisition, it is often the case that there exists a large difference between the spectral bands of satellite images collected from different geographic locations. The large shift between the spectral distributions of training and test data causes the current state-of-the-art supervised learning approaches to output unsatisfactory maps. We present a novel semantic segmentation framework that is robust to such a shift. The key component of the proposed framework is color mapping generative adversarial networks (ColorMapGANs) that can generate fake training images that are semantically exactly the same as training images, but whose spectral distribution is similar to the distribution of the test images. We then use the fake images and the ground truth for the training images to fine-tune the already trained classifier. Contrary to the existing generative adversarial networks (GANs), the generator in ColorMapGAN does not have any convolutional or pooling layers. It learns to transform the colors of the training data to the colors of the test data by performing only one elementwise matrix multiplication and one matrix-addition operation. Due to the architecturally simple but powerful design of ColorMapGAN, the proposed framework outperforms the existing approaches with a large margin in terms of both accuracy and computational complexity.

In this work, we propose a novel multi-task framework, to learn satellite image pansharpening and segmentation jointly. Our framework is based on the encoder-decoder architecture, where both tasks share the same encoder but each one has its own decoder. We compare our framework against single-task models with different architectures. Results show that our framework outperforms all other approaches in both tasks.

In dense labeling problem, the major drawback of the convolutional neural networks is their inability to learn new classes without affecting performance for the old classes on the data, having no annotations for the previous classes. In this work, we address the issue of adding new classes continually to the already trained network from a stream of data. Our approach comprises two main components: adaptation and remembering. For adaptation, we keep a clone of the previously trained network, which serves as a memory for the old classes in absence of their annotations on the new data. The updated network learns new as well as old classes on the current data using output of the memory network and the new ground-truth. For remembering, we store a little portion of the previous data, from which we systematically feed samples to the updated network during training. Our results prove that segmentation capabilities for the new classes can be added to the already trained network without catastrophically forgetting the previously learned information.

One of the most popular and challenging tasks in remote sensing applications is the generation of digitized representations of Earth's objects from satellite raster image data. A common approach to tackle this challenge is a two-step method that first involves performing a pixel-wise classification of the raster data, then vectorizing the obtained classification map. We propose a novel approach, which recasts the polygonization problem as a mesh-based approximation of the input classification map, where binary labels are assigned to the mesh triangles to represent the building class. A dense initial mesh is decimated and optimized using local edge and vertex-based operators in order to minimize an objective function that models a balance between fidelity to the classification map in l1 norm sense, right angle regularity for polygonized buildings, and final mesh complexity. Experiments show that adding the right angle objective yields better representations quantitatively and qualitatively than previous work and commonly used polygon generalization methods in remote sensing literature for similar number of vertices.

Fusion of aerial optical and LiDAR data has been a popular problem in remote sensing as they carry complementary information for object detection. We describe a stratified method that involves separately thresholding the normalized digital surface model derived from LiDAR data and the normalized difference vegetation index derived from spectral bands to obtain candidate image parts that contain different object classes, and incorporates spectral and height data with spatial information in a graph cut framework to segment the rest of the image where such separation is not possible. Experiments using a benchmark data set show that the performance of the proposed method that uses small amount of supervision is compatible with the ones in the literature.