Despite the impressive results of arbitrary image-guided style transfer methods, text-driven image stylization has recently been proposed for transferring a natural image into the stylized one according to textual descriptions of the target style provided by the user. Unlike previous image-to-image transfer approaches, text-guided stylization progress provides users with a more precise and intuitive way to express the desired style. However, the huge discrepancy between cross-modal inputs/outputs makes it challenging to conduct text-driven image stylization in a typical feed-forward CNN pipeline. In this paper, we present DiffStyler on the basis of diffusion models. The cross-modal style information can be easily integrated as guidance during the diffusion progress step-by-step. In particular, we use a dual diffusion processing architecture to control the balance between the content and style of the diffused results. Furthermore, we propose a content image-based learnable noise on which the reverse denoising process is based, enabling the stylization results to better preserve the structure information of the content image. We validate the proposed DiffStyler beyond the baseline methods through extensive qualitative and quantitative experiments.
Pre-trained Vision-Language Models (VLMs) such as CLIP have shown impressive generalization capability in downstream vision tasks with appropriate text prompts. Instead of designing prompts manually, Context Optimization (CoOp) has been recently proposed to learn continuous prompts using task-specific training data. Despite the performance improvements on downstream tasks, several studies have reported that CoOp suffers from the overfitting issue in two aspects: (i) the test accuracy on base classes first gets better and then gets worse during training; (ii) the test accuracy on novel classes keeps decreasing. However, none of the existing studies can understand and mitigate such overfitting problem effectively. In this paper, we first explore the cause of overfitting by analyzing the gradient flow. Comparative experiments reveal that CoOp favors generalizable and spurious features in the early and later training stages respectively, leading to the non-overfitting and overfitting phenomenon. Given those observations, we propose Subspace Prompt Tuning (SubPT) to project the gradients in back-propagation onto the low-rank subspace spanned by the early-stage gradient flow eigenvectors during the entire training process, and successfully eliminate the overfitting problem. Besides, we equip CoOp with Novel Feature Learner (NFL) to enhance the generalization ability of the learned prompts onto novel categories beyond the training set, needless of image training data. Extensive experiments on 11 classification datasets demonstrate that SubPT+NFL consistently boost the performance of CoOp and outperform the state-of-the-art approach CoCoOp. Experiments on more challenging vision downstream tasks including open-vocabulary object detection and zero-shot semantic segmentation also verify the effectiveness of the proposed method. Codes can be found at https://tinyurl.com/mpe64f89.
Digital art synthesis is receiving increasing attention in the multimedia community because of engaging the public with art effectively. Current digital art synthesis methods usually use single-modality inputs as guidance, thereby limiting the expressiveness of the model and the diversity of generated results. To solve this problem, we propose the multimodal guided artwork diffusion (MGAD) model, which is a diffusion-based digital artwork generation approach that utilizes multimodal prompts as guidance to control the classifier-free diffusion model. Additionally, the contrastive language-image pretraining (CLIP) model is used to unify text and image modalities. Extensive experimental results on the quality and quantity of the generated digital art paintings confirm the effectiveness of the combination of the diffusion model and multimodal guidance. Code is available at https://github.com/haha-lisa/MGAD-multimodal-guided-artwork-diffusion.
In this work, we tackle the challenging problem of arbitrary image style transfer using a novel style feature representation learning method. A suitable style representation, as a key component in image stylization tasks, is essential to achieve satisfactory results. Existing deep neural network based approaches achieve reasonable results with the guidance from second-order statistics such as Gram matrix of content features. However, they do not leverage sufficient style information, which results in artifacts such as local distortions and style inconsistency. To address these issues, we propose to learn style representation directly from image features instead of their second-order statistics, by analyzing the similarities and differences between multiple styles and considering the style distribution. Specifically, we present Contrastive Arbitrary Style Transfer (CAST), which is a new style representation learning and style transfer method via contrastive learning. Our framework consists of three key components, i.e., a multi-layer style projector for style code encoding, a domain enhancement module for effective learning of style distribution, and a generative network for image style transfer. We conduct qualitative and quantitative evaluations comprehensively to demonstrate that our approach achieves significantly better results compared to those obtained via state-of-the-art methods. Code and models are available at https://github.com/zyxElsa/CAST_pytorch
Batch normalization (BN) is widely used in modern deep neural networks, which has been shown to represent the domain-related knowledge, and thus is ineffective for cross-domain tasks like unsupervised domain adaptation (UDA). Existing BN variant methods aggregate source and target domain knowledge in the same channel in normalization module. However, the misalignment between the features of corresponding channels across domains often leads to a sub-optimal transferability. In this paper, we exploit the cross-domain relation and propose a novel normalization method, Reciprocal Normalization (RN). Specifically, RN first presents a Reciprocal Compensation (RC) module to acquire the compensatory for each channel in both domains based on the cross-domain channel-wise correlation. Then RN develops a Reciprocal Aggregation (RA) module to adaptively aggregate the feature with its cross-domain compensatory components. As an alternative to BN, RN is more suitable for UDA problems and can be easily integrated into popular domain adaptation methods. Experiments show that the proposed RN outperforms existing normalization counterparts by a large margin and helps state-of-the-art adaptation approaches achieve better results. The source code is available on https://github.com/Openning07/reciprocal-normalization-for-DA.
Vision transformers have recently received explosive popularity, but the huge computational cost is still a severe issue. Recent efficient designs for vision transformers follow two pipelines, namely, structural compression based on local spatial prior and non-structural token pruning. However, token pruning breaks the spatial structure that is indispensable for local spatial prior. To take advantage of both two pipelines, this work seeks to dynamically identify uninformative tokens for each instance and trim down both the training and inference complexity while maintaining complete spatial structure and information flow. To achieve this goal, we propose Evo-ViT, a self-motivated slow-fast token evolution method for vision transformers. Specifically, we conduct unstructured instance-wise token selection by taking advantage of the global class attention that is unique to vision transformers. Then, we propose to update informative tokens and placeholder tokens that contribute little to the final prediction with different computational priorities, namely, slow-fast updating. Thanks to the slow-fast updating mechanism that guarantees information flow and spatial structure, our Evo-ViT can accelerate vanilla transformers of both flat and deep-narrow structures from the very beginning of the training process. Experimental results demonstrate that the proposed method can significantly reduce the computational costs of vision transformers while maintaining comparable performance on image classification. For example, our method accelerates DeiTS by over 60% throughput while only sacrificing 0.4% top-1 accuracy.
Personalized image aesthetic assessment (PIAA) has recently become a hot topic due to its usefulness in a wide variety of applications such as photography, film and television, e-commerce, fashion design and so on. This task is more seriously affected by subjective factors and samples provided by users. In order to acquire precise personalized aesthetic distribution by small amount of samples, we propose a novel user-guided personalized image aesthetic assessment framework. This framework leverages user interactions to retouch and rank images for aesthetic assessment based on deep reinforcement learning (DRL), and generates personalized aesthetic distribution that is more in line with the aesthetic preferences of different users. It mainly consists of two stages. In the first stage, personalized aesthetic ranking is generated by interactive image enhancement and manual ranking, meanwhile two policy networks will be trained. The images will be pushed to the user for manual retouching and simultaneously to the enhancement policy network. The enhancement network utilizes the manual retouching results as the optimization goals of DRL. After that, the ranking process performs the similar operations like the retouching mentioned before. These two networks will be trained iteratively and alternatively to help to complete the final personalized aesthetic assessment automatically. In the second stage, these modified images are labeled with aesthetic attributes by one style-specific classifier, and then the personalized aesthetic distribution is generated based on the multiple aesthetic attributes of these images, which conforms to the aesthetic preference of users better.
The goal of image style transfer is to render an image with artistic features guided by a style reference while maintaining the original content. Due to the locality and spatial invariance in CNNs, it is difficult to extract and maintain the global information of input images. Therefore, traditional neural style transfer methods are usually biased and content leak can be observed by running several times of the style transfer process with the same reference style image. To address this critical issue, we take long-range dependencies of input images into account for unbiased style transfer by proposing a transformer-based approach, namely StyTr^2. In contrast with visual transformers for other vision tasks, our StyTr^2 contains two different transformer encoders to generate domain-specific sequences for content and style, respectively. Following the encoders, a multi-layer transformer decoder is adopted to stylize the content sequence according to the style sequence. In addition, we analyze the deficiency of existing positional encoding methods and propose the content-aware positional encoding (CAPE) which is scale-invariant and more suitable for image style transfer task. Qualitative and quantitative experiments demonstrate the effectiveness of the proposed StyTr^2 compared to state-of-the-art CNN-based and flow-based approaches.