Analyzing Neural Network-Based Generative Diffusion Models through Convex Optimization

Diffusion models are becoming widely used in state-of-the-art image, video and audio generation. Score-based diffusion models stand out among these methods, necessitating the estimation of score function of the input data distribution. In this study, we present a theoretical framework to analyze two-layer neural network-based diffusion models by reframing score matching and denoising score matching as convex optimization. Though existing diffusion theory is mainly asymptotic, we characterize the exact predicted score function and establish the convergence result for neural network-based diffusion models with finite data. This work contributes to understanding what neural network-based diffusion model learns in non-asymptotic settings.

Denoising Diffusion Probabilistic Models in Six Simple Steps

Denoising Diffusion Probabilistic Models (DDPMs) are a very popular class of deep generative model that have been successfully applied to a diverse range of problems including image and video generation, protein and material synthesis, weather forecasting, and neural surrogates of partial differential equations. Despite their ubiquity it is hard to find an introduction to DDPMs which is simple, comprehensive, clean and clear. The compact explanations necessary in research papers are not able to elucidate all of the different design steps taken to formulate the DDPM and the rationale of the steps that are presented is often omitted to save space. Moreover, the expositions are typically presented from the variational lower bound perspective which is unnecessary and arguably harmful as it obfuscates why the method is working and suggests generalisations that do not perform well in practice. On the other hand, perspectives that take the continuous time-limit are beautiful and general, but they have a high barrier-to-entry as they require background knowledge of stochastic differential equations and probability flow. In this note, we distill down the formulation of the DDPM into six simple steps each of which comes with a clear rationale. We assume that the reader is familiar with fundamental topics in machine learning including basic probabilistic modelling, Gaussian distributions, maximum likelihood estimation, and deep learning.

GaMeS: Mesh-Based Adapting and Modification of Gaussian Splatting

In recent years, a range of neural network-based methods for image rendering have been introduced. For instance, widely-researched neural radiance fields (NeRF) rely on a neural network to represent 3D scenes, allowing for realistic view synthesis from a small number of 2D images. However, most NeRF models are constrained by long training and inference times. In comparison, Gaussian Splatting (GS) is a novel, state-of-theart technique for rendering points in a 3D scene by approximating their contribution to image pixels through Gaussian distributions, warranting fast training and swift, real-time rendering. A drawback of GS is the absence of a well-defined approach for its conditioning due to the necessity to condition several hundred thousand Gaussian components. To solve this, we introduce Gaussian Mesh Splatting (GaMeS) model, a hybrid of mesh and a Gaussian distribution, that pin all Gaussians splats on the object surface (mesh). The unique contribution of our methods is defining Gaussian splats solely based on their location on the mesh, allowing for automatic adjustments in position, scale, and rotation during animation. As a result, we obtain high-quality renders in the real-time generation of high-quality views. Furthermore, we demonstrate that in the absence of a predefined mesh, it is possible to fine-tune the initial mesh during the learning process.

AmbientCycleGAN for Establishing Interpretable Stochastic Object Models Based on Mathematical Phantoms and Medical Imaging Measurements

Medical imaging systems that are designed for producing diagnostically informative images should be objectively assessed via task-based measures of image quality (IQ). Ideally, computation of task-based measures of IQ needs to account for all sources of randomness in the measurement data, including the variability in the ensemble of objects to be imaged. To address this need, stochastic object models (SOMs) that can generate an ensemble of synthesized objects or phantoms can be employed. Various mathematical SOMs or phantoms were developed that can interpretably synthesize objects, such as lumpy object models and parameterized torso phantoms. However, such SOMs that are purely mathematically defined may not be able to comprehensively capture realistic object variations. To establish realistic SOMs, it is desirable to use experimental data. An augmented generative adversarial network (GAN), AmbientGAN, was recently proposed for establishing SOMs from medical imaging measurements. However, it remains unclear to which extent the AmbientGAN-produced objects can be interpretably controlled. This work introduces a novel approach called AmbientCycleGAN that translates mathematical SOMs to realistic SOMs by use of noisy measurement data. Numerical studies that consider clustered lumpy background (CLB) models and real mammograms are conducted. It is demonstrated that our proposed method can stably establish SOMs based on mathematical models and noisy measurement data. Moreover, the ability of the proposed AmbientCycleGAN to interpretably control image features in the synthesized objects is investigated.

Jailbreaking Attack against Multimodal Large Language Model

This paper focuses on jailbreaking attacks against multi-modal large language models (MLLMs), seeking to elicit MLLMs to generate objectionable responses to harmful user queries. A maximum likelihood-based algorithm is proposed to find an \emph{image Jailbreaking Prompt} (imgJP), enabling jailbreaks against MLLMs across multiple unseen prompts and images (i.e., data-universal property). Our approach exhibits strong model-transferability, as the generated imgJP can be transferred to jailbreak various models, including MiniGPT-v2, LLaVA, InstructBLIP, and mPLUG-Owl2, in a black-box manner. Moreover, we reveal a connection between MLLM-jailbreaks and LLM-jailbreaks. As a result, we introduce a construction-based method to harness our approach for LLM-jailbreaks, demonstrating greater efficiency than current state-of-the-art methods. The code is available here. \textbf{Warning: some content generated by language models may be offensive to some readers.}

SCoFT: Self-Contrastive Fine-Tuning for Equitable Image Generation

Accurate representation in media is known to improve the well-being of the people who consume it. Generative image models trained on large web-crawled datasets such as LAION are known to produce images with harmful stereotypes and misrepresentations of cultures. We improve inclusive representation in generated images by (1) engaging with communities to collect a culturally representative dataset that we call the Cross-Cultural Understanding Benchmark (CCUB) and (2) proposing a novel Self-Contrastive Fine-Tuning (SCoFT) method that leverages the model's known biases to self-improve. SCoFT is designed to prevent overfitting on small datasets, encode only high-level information from the data, and shift the generated distribution away from misrepresentations encoded in a pretrained model. Our user study conducted on 51 participants from 5 different countries based on their self-selected national cultural affiliation shows that fine-tuning on CCUB consistently generates images with higher cultural relevance and fewer stereotypes when compared to the Stable Diffusion baseline, which is further improved with our SCoFT technique.

StableIdentity: Inserting Anybody into Anywhere at First Sight

Recent advances in large pretrained text-to-image models have shown unprecedented capabilities for high-quality human-centric generation, however, customizing face identity is still an intractable problem. Existing methods cannot ensure stable identity preservation and flexible editability, even with several images for each subject during training. In this work, we propose StableIdentity, which allows identity-consistent recontextualization with just one face image. More specifically, we employ a face encoder with an identity prior to encode the input face, and then land the face representation into a space with an editable prior, which is constructed from celeb names. By incorporating identity prior and editability prior, the learned identity can be injected anywhere with various contexts. In addition, we design a masked two-phase diffusion loss to boost the pixel-level perception of the input face and maintain the diversity of generation. Extensive experiments demonstrate our method outperforms previous customization methods. In addition, the learned identity can be flexibly combined with the off-the-shelf modules such as ControlNet. Notably, to the best knowledge, we are the first to directly inject the identity learned from a single image into video/3D generation without finetuning. We believe that the proposed StableIdentity is an important step to unify image, video, and 3D customized generation models.

Binding Touch to Everything: Learning Unified Multimodal Tactile Representations

The ability to associate touch with other modalities has huge implications for humans and computational systems. However, multimodal learning with touch remains challenging due to the expensive data collection process and non-standardized sensor outputs. We introduce UniTouch, a unified tactile model for vision-based touch sensors connected to multiple modalities, including vision, language, and sound. We achieve this by aligning our UniTouch embeddings to pretrained image embeddings already associated with a variety of other modalities. We further propose learnable sensor-specific tokens, allowing the model to learn from a set of heterogeneous tactile sensors, all at the same time. UniTouch is capable of conducting various touch sensing tasks in the zero-shot setting, from robot grasping prediction to touch image question answering. To the best of our knowledge, UniTouch is the first to demonstrate such capabilities. Project page: https://cfeng16.github.io/UniTouch/

Selection of gamma events from IACT images with deep learning methods

Imaging Atmospheric Cherenkov Telescopes (IACTs) of gamma ray observatory TAIGA detect the Extesnive Air Showers (EASs) originating from the cosmic or gamma rays interactions with the atmosphere. Thereby, telescopes obtain images of the EASs. The ability to segregate gamma rays images from the hadronic cosmic ray background is one of the main features of this type of detectors. However, in actual IACT observations simultaneous observation of the background and the source of gamma ray is needed. This observation mode (called wobbling) modifies images of events, which affects the quality of selection by neural networks. Thus, in this work, the results of the application of neural networks (NN) for image classification task on Monte Carlo (MC) images of TAIGA-IACTs are presented. The wobbling mode is considered together with the image adaptation for adequate analysis by NNs. Simultaneously, we explore several neural network structures that classify events both directly from images or through Hillas parameters extracted from images. In addition, by employing NNs, MC simulation data are used to evaluate the quality of the segregation of rare gamma events with the account of all necessary image modifications.

Towards Precise 3D Human Pose Estimation with Multi-Perspective Spatial-Temporal Relational Transformers

3D human pose estimation captures the human joint points in three-dimensional space while keeping the depth information and physical structure. That is essential for applications that require precise pose information, such as human-computer interaction, scene understanding, and rehabilitation training. Due to the challenges in data collection, mainstream datasets of 3D human pose estimation are primarily composed of multi-view video data collected in laboratory environments, which contains rich spatial-temporal correlation information besides the image frame content. Given the remarkable self-attention mechanism of transformers, capable of capturing the spatial-temporal correlation from multi-view video datasets, we propose a multi-stage framework for 3D sequence-to-sequence (seq2seq) human pose detection. Firstly, the spatial module represents the human pose feature by intra-image content, while the frame-image relation module extracts temporal relationships and 3D spatial positional relationship features between the multi-perspective images. Secondly, the self-attention mechanism is adopted to eliminate the interference from non-human body parts and reduce computing resources. Our method is evaluated on Human3.6M, a popular 3D human pose detection dataset. Experimental results demonstrate that our approach achieves state-of-the-art performance on this dataset.