Vision-language models (VLMs) such as CLIP are trained via contrastive learning between text and image pairs, resulting in aligned image and text embeddings that are useful for many downstream tasks. A notable drawback of CLIP, however, is that the resulting embedding space seems to lack some of the structure of their purely text-based alternatives. For instance, while text embeddings have been long noted to satisfy \emph{analogies} in embedding space using vector arithmetic, CLIP has no such property. In this paper, we propose an approach to natively train CLIP in a contrastive manner to reason about differences in embedding space. We finetune CLIP so that the differences in image embedding space correspond to \emph{text descriptions of the image differences}, which we synthetically generate with large language models on image-caption paired datasets. We first demonstrate that our approach yields significantly improved capabilities in ranking images by a certain attribute (e.g., elephants are larger than cats), which is useful in retrieval or constructing attribute-based classifiers, and improved zeroshot classification performance on many downstream image classification tasks. In addition, our approach enables a new mechanism for inference that we refer to as comparative prompting, where we leverage prior knowledge of text descriptions of differences between classes of interest, achieving even larger performance gains in classification. Finally, we illustrate that the resulting embeddings obey a larger degree of geometric properties in embedding space, such as in text-to-image generation.
E-commerce image generation has always been one of the core demands in the e-commerce field. The goal is to restore the missing background that matches the main product given. In the post-AIGC era, diffusion models are primarily used to generate product images, achieving impressive results. This paper systematically analyzes and addresses a core pain point in diffusion model generation: overcompletion, which refers to the difficulty in maintaining product features. We propose two solutions: 1. Using an instance mask fine-tuned inpainting model to mitigate this phenomenon; 2. Adopting a train-free mask guidance approach, which incorporates refined product masks as constraints when combining ControlNet and UNet to generate the main product, thereby avoiding overcompletion of the product. Our method has achieved promising results in practical applications and we hope it can serve as an inspiring technical report in this field.
Current Pose-Guided Person Image Synthesis (PGPIS) methods depend heavily on large amounts of labeled triplet data to train the generator in a supervised manner. However, they often falter when applied to in-the-wild samples, primarily due to the distribution gap between the training datasets and real-world test samples. While some researchers aim to enhance model generalizability through sophisticated training procedures, advanced architectures, or by creating more diverse datasets, we adopt the test-time fine-tuning paradigm to customize a pre-trained Text2Image (T2I) model. However, naively applying test-time tuning results in inconsistencies in facial identities and appearance attributes. To address this, we introduce a Visual Consistency Module (VCM), which enhances appearance consistency by combining the face, text, and image embedding. Our approach, named OnePoseTrans, requires only a single source image to generate high-quality pose transfer results, offering greater stability than state-of-the-art data-driven methods. For each test case, OnePoseTrans customizes a model in around 48 seconds with an NVIDIA V100 GPU.
Contrastive pretraining can substantially increase model generalisation and downstream performance. However, the quality of the learned representations is highly dependent on the data augmentation strategy applied to generate positive pairs. Positive contrastive pairs should preserve semantic meaning while discarding unwanted variations related to the data acquisition domain. Traditional contrastive pipelines attempt to simulate domain shifts through pre-defined generic image transformations. However, these do not always mimic realistic and relevant domain variations for medical imaging such as scanner differences. To tackle this issue, we herein introduce counterfactual contrastive learning, a novel framework leveraging recent advances in causal image synthesis to create contrastive positive pairs that faithfully capture relevant domain variations. Our method, evaluated across five datasets encompassing both chest radiography and mammography data, for two established contrastive objectives (SimCLR and DINO-v2), outperforms standard contrastive learning in terms of robustness to acquisition shift. Notably, counterfactual contrastive learning achieves superior downstream performance on both in-distribution and on external datasets, especially for images acquired with scanners under-represented in the training set. Further experiments show that the proposed framework extends beyond acquisition shifts, with models trained with counterfactual contrastive learning substantially improving subgroup performance across biological sex.
The influence of textures on machine learning models has been an ongoing investigation, specifically in texture bias/learning, interpretability, and robustness. However, due to the lack of large and diverse texture data available, the findings in these works have been limited, as more comprehensive evaluations have not been feasible. Image generative models are able to provide data creation at scale, but utilizing these models for texture synthesis has been unexplored and poses additional challenges both in creating accurate texture images and validating those images. In this work, we introduce an extensible methodology and corresponding new dataset for generating high-quality, diverse texture images capable of supporting a broad set of texture-based tasks. Our pipeline consists of: (1) developing prompts from a range of descriptors to serve as input to text-to-image models, (2) adopting and adapting Stable Diffusion pipelines to generate and filter the corresponding images, and (3) further filtering down to the highest quality images. Through this, we create the Prompted Textures Dataset (PTD), a dataset of 246,285 texture images that span 56 textures. During the process of generating images, we find that NSFW safety filters in image generation pipelines are highly sensitive to texture (and flag up to 60\% of our texture images), uncovering a potential bias in these models and presenting unique challenges when working with texture data. Through both standard metrics and a human evaluation, we find that our dataset is high quality and diverse. Our dataset is available for download at https://zenodo.org/records/15359142.
Recent studies in text-to-image customization show great success in generating personalized object variants given several images of a subject. While existing methods focus more on preserving the identity of the subject, they often fall short of controlling the spatial relationship between objects. In this work, we introduce GroundingBooth, a framework that achieves zero-shot instance-level spatial grounding on both foreground subjects and background objects in the text-to-image customization task. Our proposed text-image grounding module and masked cross-attention layer allow us to generate personalized images with both accurate layout alignment and identity preservation while maintaining text-image coherence. With such layout control, our model inherently enables the customization of multiple subjects at once. Our model is evaluated on both layout-guided image synthesis and reference-based customization tasks, showing strong results compared to existing methods. Our work is the first work to achieve a joint grounding on both subject-driven foreground generation and text-driven background generation.
With the emerging trend of large generative models, ControlNet is introduced to enable users to fine-tune pre-trained models with their own data for various use cases. A natural question arises: how can we train ControlNet models while ensuring users' data privacy across distributed devices? Exploring different distributed training schemes, we find conventional federated learning and split learning unsuitable. Instead, we propose a new distributed learning structure that eliminates the need for the server to send gradients back. Through a comprehensive evaluation of existing threats, we discover that in the context of training ControlNet with split learning, most existing attacks are ineffective, except for two mentioned in previous literature. To counter these threats, we leverage the properties of diffusion models and design a new timestep sampling policy during forward processes. We further propose a privacy-preserving activation function and a method to prevent private text prompts from leaving clients, tailored for image generation with diffusion models. Our experimental results demonstrate that our algorithms and systems greatly enhance the efficiency of distributed training for ControlNet while ensuring users' data privacy without compromising image generation quality.
Drag-based image editing has recently gained popularity for its interactivity and precision. However, despite the ability of text-to-image models to generate samples within a second, drag editing still lags behind due to the challenge of accurately reflecting user interaction while maintaining image content. Some existing approaches rely on computationally intensive per-image optimization or intricate guidance-based methods, requiring additional inputs such as masks for movable regions and text prompts, thereby compromising the interactivity of the editing process. We introduce InstantDrag, an optimization-free pipeline that enhances interactivity and speed, requiring only an image and a drag instruction as input. InstantDrag consists of two carefully designed networks: a drag-conditioned optical flow generator (FlowGen) and an optical flow-conditioned diffusion model (FlowDiffusion). InstantDrag learns motion dynamics for drag-based image editing in real-world video datasets by decomposing the task into motion generation and motion-conditioned image generation. We demonstrate InstantDrag's capability to perform fast, photo-realistic edits without masks or text prompts through experiments on facial video datasets and general scenes. These results highlight the efficiency of our approach in handling drag-based image editing, making it a promising solution for interactive, real-time applications.
Today's image generation systems are capable of producing realistic and high-quality images. However, user prompts often contain ambiguities, making it difficult for these systems to interpret users' potential intentions. Consequently, machines need to interact with users multiple rounds to better understand users' intents. The unpredictable costs of using or learning image generation models through multiple feedback interactions hinder their widespread adoption and full performance potential, especially for non-expert users. In this research, we aim to enhance the user-friendliness of our image generation system. To achieve this, we propose a reflective human-machine co-adaptation strategy, named RHM-CAS. Externally, the Agent engages in meaningful language interactions with users to reflect on and refine the generated images. Internally, the Agent tries to optimize the policy based on user preferences, ensuring that the final outcomes closely align with user preferences. Various experiments on different tasks demonstrate the effectiveness of the proposed method.
The porous media community extensively utilizes digital rock images for core analysis. High-resolution digital rock images that possess sufficient quality are essential but often challenging to acquire. Super-resolution (SR) approaches enhance the resolution of digital rock images and provide improved visualization of fine features and structures, aiding in the analysis and interpretation of rock properties, such as pore connectivity and mineral distribution. However, there is a current shortage of real paired microscopic images for super-resolution training. In this study, we used two types of Scanning Electron Microscopes (SEM) to obtain the images of shale samples in five regions, with 1X, 2X, 4X, 8X and 16X magnifications. We used these real scanned paired images as a reference to select the optimal method of image generation and validated it using Enhanced Deep Super Resolution (EDSR) and Very Deep Super Resolution (VDSR) methods. Our experiments show that the bilinear algorithm is more suitable than the commonly used bicubic method, for establishing low-resolution datasets in the SR approaches, which is partially attributed to the mechanism of Scanning Electron Microscopes (SEM).
This article introduces Bio-Eng-LMM AI chatbot, a versatile platform designed to enhance user interaction for educational and research purposes. Leveraging cutting-edge open-source Large Language Models (LLMs), Bio-Eng-LMM operates as a sophisticated AI assistant, exploiting the capabilities of traditional models like ChatGPT. Central to Bio-Eng-LMM is its implementation of Retrieval Augmented Generation (RAG) through three primary methods: integration of preprocessed documents, real-time processing of user-uploaded files, and information retrieval from any specified website. Additionally, the chatbot incorporates image generation via a Stable Diffusion Model (SDM), image understanding and response generation through LLAVA, and search functionality on the internet powered by secure search engine such as DuckDuckGo. To provide comprehensive support, Bio-Eng-LMM offers text summarization, website content summarization, and both text and voice interaction. The chatbot maintains session memory to ensure contextually relevant and coherent responses. This integrated platform builds upon the strengths of RAG-GPT and Web-Based RAG Query (WBRQ) where the system fetches relevant information directly from the web to enhance the LLMs response generation.
Recently, text-to-image generative models have been misused to create unauthorized malicious images of individuals, posing a growing social problem. Previous solutions, such as Anti-DreamBooth, add adversarial noise to images to protect them from being used as training data for malicious generation. However, we found that the adversarial noise can be removed by adversarial purification methods such as DiffPure. Therefore, we propose a new adversarial attack method that adds strong perturbation on the high-frequency areas of images to make it more robust to adversarial purification. Our experiment showed that the adversarial images retained noise even after adversarial purification, hindering malicious image generation.
Zero-shot personalized image generation models aim to produce images that align with both a given text prompt and subject image, requiring the model to effectively incorporate both sources of guidance. However, existing methods often struggle to capture fine-grained subject details and frequently prioritize one form of guidance over the other, resulting in suboptimal subject encoding and an imbalance in the generated images. In this study, we uncover key insights into achieving high-quality balances on subject identity preservation and text-following, notably that 1) the design of the subject image encoder critically influences subject identity preservation, and 2) the text and subject guidance should take effect at different denoising stages. Building on these insights, we introduce a new approach, EZIGen, that employs two main components: a carefully crafted subject image encoder based on the pre-trained UNet of the Stable Diffusion model, following a process that balances the two guidances by separating their dominance stage and revisiting certain time steps to bootstrap subject transfer quality. Through these two components, EZIGen achieves state-of-the-art results on multiple personalized generation benchmarks with a unified model and 100 times less training data. Demo Page: zichengduan.github.io/pages/EZIGen/index.html
Recent advancements in text-to-image diffusion models have demonstrated remarkable success, yet they often struggle to fully capture the user's intent. Existing approaches using textual inputs combined with bounding boxes or region masks fall short in providing precise spatial guidance, often leading to misaligned or unintended object orientation. To address these limitations, we propose Scribble-Guided Diffusion (ScribbleDiff), a training-free approach that utilizes simple user-provided scribbles as visual prompts to guide image generation. However, incorporating scribbles into diffusion models presents challenges due to their sparse and thin nature, making it difficult to ensure accurate orientation alignment. To overcome these challenges, we introduce moment alignment and scribble propagation, which allow for more effective and flexible alignment between generated images and scribble inputs. Experimental results on the PASCAL-Scribble dataset demonstrate significant improvements in spatial control and consistency, showcasing the effectiveness of scribble-based guidance in diffusion models. Our code is available at https://github.com/kaist-cvml-lab/scribble-diffusion.
Diffusion Transformers (DiTs) have recently attracted significant interest from both industry and academia due to their enhanced capabilities in visual generation, surpassing the performance of traditional diffusion models that employ U-Net. However, the improved performance of DiTs comes at the expense of higher parameter counts and implementation costs, which significantly limits their deployment on resource-constrained devices like mobile phones. We propose DiTAS, a data-free post-training quantization (PTQ) method for efficient DiT inference. DiTAS relies on the proposed temporal-aggregated smoothing techniques to mitigate the impact of the channel-wise outliers within the input activations, leading to much lower quantization error under extremely low bitwidth. To further enhance the performance of the quantized DiT, we adopt the layer-wise grid search strategy to optimize the smoothing factor. Moreover, we integrate a training-free LoRA module for weight quantization, leveraging alternating optimization to minimize quantization errors without additional fine-tuning. Experimental results demonstrate that our approach enables 4-bit weight, 8-bit activation (W4A8) quantization for DiTs while maintaining comparable performance as the full-precision model.
Initially designed for image generation, diffusion models can also be effectively applied to various tasks, including semantic segmentation. However, most existing diffusion-based approaches for semantic segmentation operate in high-dimensional pixel space, demanding a lot of computing and memory resources during training and inference. This paper makes the first attempt to utilize latent diffusion models for semantic segmentation. Specifically, we propose a fast yet effective latent diffusion model and valuate it on medical image segmentation tasks. Firstly, we train a Variational Autoencoder (VAE) network to convert binary image masks into compact latent vectors. The diffusion process can then be executed in this low-dimensional latent space and thus drastically accelerated. Subsequently, we employ the VAE decoder to reconstruct a precise prediction map from the latent output vector produced by the diffusion process. Eventually, we refine the final segmentation results through a straightforward post-processing step using morphological operations. We report our results on two public datasets, including colon polyp images and skin cancer images. Experiments show that our approach achieves competitive accuracy compared to traditional diffusion models while having much better training and inference speed, as well as much more efficient memory consumption.
Recent advancements in generative models have revolutionized image generation and editing, making these tasks accessible to non-experts. This paper focuses on local image editing, particularly the task of adding new content to a loosely specified area. Existing methods often require a precise mask or a detailed description of the location, which can be cumbersome and prone to errors. We propose Click2Mask, a novel approach that simplifies the local editing process by requiring only a single point of reference (in addition to the content description). A mask is dynamically grown around this point during a Blended Latent Diffusion (BLD) process, guided by a masked CLIP-based semantic loss. Click2Mask surpasses the limitations of segmentation-based and fine-tuning dependent methods, offering a more user-friendly and contextually accurate solution. Our experiments demonstrate that Click2Mask not only minimizes user effort but also enables competitive or superior local image manipulations compared to SoTA methods, according to both human judgement and automatic metrics. Key contributions include the simplification of user input, the ability to freely add objects unconstrained by existing segments, and the integration potential of our dynamic mask approach within other editing methods.
Diffusion models have led to the revolutionizing of generative modeling in numerous image synthesis tasks. Nevertheless, it is not trivial to directly apply diffusion models for synthesizing an image of a target person wearing a given in-shop garment, i.e., image-based virtual try-on (VTON) task. The difficulty originates from the aspect that the diffusion process should not only produce holistically high-fidelity photorealistic image of the target person, but also locally preserve every appearance and texture detail of the given garment. To address this, we shape a new Diffusion model, namely GarDiff, which triggers the garment-focused diffusion process with amplified guidance of both basic visual appearance and detailed textures (i.e., high-frequency details) derived from the given garment. GarDiff first remoulds a pre-trained latent diffusion model with additional appearance priors derived from the CLIP and VAE encodings of the reference garment. Meanwhile, a novel garment-focused adapter is integrated into the UNet of diffusion model, pursuing local fine-grained alignment with the visual appearance of reference garment and human pose. We specifically design an appearance loss over the synthesized garment to enhance the crucial, high-frequency details. Extensive experiments on VITON-HD and DressCode datasets demonstrate the superiority of our GarDiff when compared to state-of-the-art VTON approaches. Code is publicly available at: \href{https://github.com/siqi0905/GarDiff/tree/master}{https://github.com/siqi0905/GarDiff/tree/master}.
Recent breakthroughs in text-to-image models have opened up promising research avenues in personalized image generation, enabling users to create diverse images of a specific subject using natural language prompts. However, existing methods often suffer from performance degradation when given only a single reference image. They tend to overfit the input, producing highly similar outputs regardless of the text prompt. This paper addresses the challenge of one-shot personalization by mitigating overfitting, enabling the creation of controllable images through text prompts. Specifically, we propose a selective fine-tuning strategy that focuses on the text encoder. Furthermore, we introduce three key techniques to enhance personalization performance: (1) augmentation tokens to encourage feature disentanglement and alleviate overfitting, (2) a knowledge-preservation loss to reduce language drift and promote generalizability across diverse prompts, and (3) SNR-weighted sampling for efficient training. Extensive experiments demonstrate that our approach efficiently generates high-quality, diverse images using only a single reference image while significantly reducing memory and storage requirements.
While Text-to-Image (T2I) diffusion models excel at generating visually appealing images of individual instances, they struggle to accurately position and control the features generation of multiple instances. The Layout-to-Image (L2I) task was introduced to address the positioning challenges by incorporating bounding boxes as spatial control signals, but it still falls short in generating precise instance features. In response, we propose the Instance Feature Generation (IFG) task, which aims to ensure both positional accuracy and feature fidelity in generated instances. To address the IFG task, we introduce the Instance Feature Adapter (IFAdapter). The IFAdapter enhances feature depiction by incorporating additional appearance tokens and utilizing an Instance Semantic Map to align instance-level features with spatial locations. The IFAdapter guides the diffusion process as a plug-and-play module, making it adaptable to various community models. For evaluation, we contribute an IFG benchmark and develop a verification pipeline to objectively compare models' abilities to generate instances with accurate positioning and features. Experimental results demonstrate that IFAdapter outperforms other models in both quantitative and qualitative evaluations.