Federated Black-Box Adaptation for Semantic Segmentation

Federated Learning (FL) is a form of distributed learning that allows multiple institutions or clients to collaboratively learn a global model to solve a task. This allows the model to utilize the information from every institute while preserving data privacy. However, recent studies show that the promise of protecting the privacy of data is not upheld by existing methods and that it is possible to recreate the training data from the different institutions. This is done by utilizing gradients transferred between the clients and the global server during training or by knowing the model architecture at the client end. In this paper, we propose a federated learning framework for semantic segmentation without knowing the model architecture nor transferring gradients between the client and the server, thus enabling better privacy preservation. We propose BlackFed - a black-box adaptation of neural networks that utilizes zero order optimization (ZOO) to update the client model weights and first order optimization (FOO) to update the server weights. We evaluate our approach on several computer vision and medical imaging datasets to demonstrate its effectiveness. To the best of our knowledge, this work is one of the first works in employing federated learning for segmentation, devoid of gradients or model information exchange. Code: https://github.com/JayParanjape/blackfed/tree/master

Active Learning for Vision-Language Models

Pre-trained vision-language models (VLMs) like CLIP have demonstrated impressive zero-shot performance on a wide range of downstream computer vision tasks. However, there still exists a considerable performance gap between these models and a supervised deep model trained on a downstream dataset. To bridge this gap, we propose a novel active learning (AL) framework that enhances the zero-shot classification performance of VLMs by selecting only a few informative samples from the unlabeled data for annotation during training. To achieve this, our approach first calibrates the predicted entropy of VLMs and then utilizes a combination of self-uncertainty and neighbor-aware uncertainty to calculate a reliable uncertainty measure for active sample selection. Our extensive experiments show that the proposed approach outperforms existing AL approaches on several image classification datasets, and significantly enhances the zero-shot performance of VLMs.

AWRaCLe: All-Weather Image Restoration using Visual In-Context Learning

All-Weather Image Restoration (AWIR) under adverse weather conditions is a challenging task due to the presence of different types of degradations. Prior research in this domain relies on extensive training data but lacks the utilization of additional contextual information for restoration guidance. Consequently, the performance of existing methods is limited by the degradation cues that are learnt from individual training samples. Recent advancements in visual in-context learning have introduced generalist models that are capable of addressing multiple computer vision tasks simultaneously by using the information present in the provided context as a prior. In this paper, we propose All-Weather Image Restoration using Visual In-Context Learning (AWRaCLe), a novel approach for AWIR that innovatively utilizes degradation-specific visual context information to steer the image restoration process. To achieve this, AWRaCLe incorporates Degradation Context Extraction (DCE) and Context Fusion (CF) to seamlessly integrate degradation-specific features from the context into an image restoration network. The proposed DCE and CF blocks leverage CLIP features and incorporate attention mechanisms to adeptly learn and fuse contextual information. These blocks are specifically designed for visual in-context learning under all-weather conditions and are crucial for effective context utilization. Through extensive experiments, we demonstrate the effectiveness of AWRaCLe for all-weather restoration and show that our method advances the state-of-the-art in AWIR.

S-SAM: SVD-based Fine-Tuning of Segment Anything Model for Medical Image Segmentation

Medical image segmentation has been traditionally approached by training or fine-tuning the entire model to cater to any new modality or dataset. However, this approach often requires tuning a large number of parameters during training. With the introduction of the Segment Anything Model (SAM) for prompted segmentation of natural images, many efforts have been made towards adapting it efficiently for medical imaging, thus reducing the training time and resources. However, these methods still require expert annotations for every image in the form of point prompts or bounding box prompts during training and inference, making it tedious to employ them in practice. In this paper, we propose an adaptation technique, called S-SAM, that only trains parameters equal to 0.4% of SAM's parameters and at the same time uses simply the label names as prompts for producing precise masks. This not only makes tuning SAM more efficient than the existing adaptation methods but also removes the burden of providing expert prompts. We call this modified version S-SAM and evaluate it on five different modalities including endoscopic images, x-ray, ultrasound, CT, and histology images. Our experiments show that S-SAM outperforms state-of-the-art methods as well as existing SAM adaptation methods while tuning a significantly less number of parameters. We release the code for S-SAM at https://github.com/JayParanjape/SVDSAM.

Dense Multimodal Alignment for Open-Vocabulary 3D Scene Understanding

Recent vision-language pre-training models have exhibited remarkable generalization ability in zero-shot recognition tasks. Previous open-vocabulary 3D scene understanding methods mostly focus on training 3D models using either image or text supervision while neglecting the collective strength of all modalities. In this work, we propose a Dense Multimodal Alignment (DMA) framework to densely co-embed different modalities into a common space for maximizing their synergistic benefits. Instead of extracting coarse view- or region-level text prompts, we leverage large vision-language models to extract complete category information and scalable scene descriptions to build the text modality, and take image modality as the bridge to build dense point-pixel-text associations. Besides, in order to enhance the generalization ability of the 2D model for downstream 3D tasks without compromising the open-vocabulary capability, we employ a dual-path integration approach to combine frozen CLIP visual features and learnable mask features. Extensive experiments show that our DMA method produces highly competitive open-vocabulary segmentation performance on various indoor and outdoor tasks.

Reference-based Controllable Scene Stylization with Gaussian Splatting

Referenced-based scene stylization that edits the appearance based on a content-aligned reference image is an emerging research area. Starting with a pretrained neural radiance field (NeRF), existing methods typically learn a novel appearance that matches the given style. Despite their effectiveness, they inherently suffer from time-consuming volume rendering, and thus are impractical for many real-time applications. In this work, we propose ReGS, which adapts 3D Gaussian Splatting (3DGS) for reference-based stylization to enable real-time stylized view synthesis. Editing the appearance of a pretrained 3DGS is challenging as it uses discrete Gaussians as 3D representation, which tightly bind appearance with geometry. Simply optimizing the appearance as prior methods do is often insufficient for modeling continuous textures in the given reference image. To address this challenge, we propose a novel texture-guided control mechanism that adaptively adjusts local responsible Gaussians to a new geometric arrangement, serving for desired texture details. The proposed process is guided by texture clues for effective appearance editing, and regularized by scene depth for preserving original geometric structure. With these novel designs, we show ReGs can produce state-of-the-art stylization results that respect the reference texture while embracing real-time rendering speed for free-view navigation.

JeDi: Joint-Image Diffusion Models for Finetuning-Free Personalized Text-to-Image Generation

Personalized text-to-image generation models enable users to create images that depict their individual possessions in diverse scenes, finding applications in various domains. To achieve the personalization capability, existing methods rely on finetuning a text-to-image foundation model on a user's custom dataset, which can be non-trivial for general users, resource-intensive, and time-consuming. Despite attempts to develop finetuning-free methods, their generation quality is much lower compared to their finetuning counterparts. In this paper, we propose Joint-Image Diffusion (\jedi), an effective technique for learning a finetuning-free personalization model. Our key idea is to learn the joint distribution of multiple related text-image pairs that share a common subject. To facilitate learning, we propose a scalable synthetic dataset generation technique. Once trained, our model enables fast and easy personalization at test time by simply using reference images as input during the sampling process. Our approach does not require any expensive optimization process or additional modules and can faithfully preserve the identity represented by any number of reference images. Experimental results show that our model achieves state-of-the-art generation quality, both quantitatively and qualitatively, significantly outperforming both the prior finetuning-based and finetuning-free personalization baselines.

A Mamba-based Siamese Network for Remote Sensing Change Detection

Change detection in remote sensing images is an essential tool for analyzing a region at different times. It finds varied applications in monitoring environmental changes, man-made changes as well as corresponding decision-making and prediction of future trends. Deep learning methods like Convolutional Neural Networks (CNNs) and Transformers have achieved remarkable success in detecting significant changes, given two images at different times. In this paper, we propose a Mamba-based Change Detector (M-CD) that segments out the regions of interest even better. Mamba-based architectures demonstrate linear-time training capabilities and an improved receptive field over transformers. Our experiments on four widely used change detection datasets demonstrate significant improvements over existing state-of-the-art (SOTA) methods. Our code and pre-trained models are available at https://github.com/JayParanjape/M-CD

SyncNoise: Geometrically Consistent Noise Prediction for Text-based 3D Scene Editing

Text-based 2D diffusion models have demonstrated impressive capabilities in image generation and editing. Meanwhile, the 2D diffusion models also exhibit substantial potentials for 3D editing tasks. However, how to achieve consistent edits across multiple viewpoints remains a challenge. While the iterative dataset update method is capable of achieving global consistency, it suffers from slow convergence and over-smoothed textures. We propose SyncNoise, a novel geometry-guided multi-view consistent noise editing approach for high-fidelity 3D scene editing. SyncNoise synchronously edits multiple views with 2D diffusion models while enforcing multi-view noise predictions to be geometrically consistent, which ensures global consistency in both semantic structure and low-frequency appearance. To further enhance local consistency in high-frequency details, we set a group of anchor views and propagate them to their neighboring frames through cross-view reprojection. To improve the reliability of multi-view correspondences, we introduce depth supervision during training to enhance the reconstruction of precise geometries. Our method achieves high-quality 3D editing results respecting the textual instructions, especially in scenes with complex textures, by enhancing geometric consistency at the noise and pixel levels.

ModelMix: A New Model-Mixup Strategy to Minimize Vicinal Risk across Tasks for Few-scribble based Cardiac Segmentation

Pixel-level dense labeling is both resource-intensive and time-consuming, whereas weak labels such as scribble present a more feasible alternative to full annotations. However, training segmentation networks with weak supervision from scribbles remains challenging. Inspired by the fact that different segmentation tasks can be correlated with each other, we introduce a new approach to few-scribble supervised segmentation based on model parameter interpolation, termed as ModelMix. Leveraging the prior knowledge that linearly interpolating convolution kernels and bias terms should result in linear interpolations of the corresponding feature vectors, ModelMix constructs virtual models using convex combinations of convolutional parameters from separate encoders. We then regularize the model set to minimize vicinal risk across tasks in both unsupervised and scribble-supervised way. Validated on three open datasets, i.e., ACDC, MSCMRseg, and MyoPS, our few-scribble guided ModelMix significantly surpasses the performance of the state-of-the-art scribble supervised methods.