This research addresses a critical challenge in the field of generative models, particularly in the generation and evaluation of synthetic images. Given the inherent complexity of generative models and the absence of a standardized procedure for their comparison, our study introduces a pioneering algorithm to objectively assess the realism of synthetic images. This approach significantly enhances the evaluation methodology by refining the Fr\'echet Inception Distance (FID) score, allowing for a more precise and subjective assessment of image quality. Our algorithm is particularly tailored to address the challenges in generating and evaluating realistic images of Arabic handwritten digits, a task that has traditionally been near-impossible due to the subjective nature of realism in image generation. By providing a systematic and objective framework, our method not only enables the comparison of different generative models but also paves the way for improvements in their design and output. This breakthrough in evaluation and comparison is crucial for advancing the field of OCR, especially for scripts that present unique complexities, and sets a new standard in the generation and assessment of high-quality synthetic images.
Achieving successful variable bitrate compression with computationally simple algorithms from a single end-to-end learned image or video compression model remains a challenge. Many approaches have been proposed, including conditional auto-encoders, channel-adaptive gains for the latent tensor or uniformly quantizing all elements of the latent tensor. This paper follows the traditional approach to vary a single quantization step size to perform uniform quantization of all latent tensor elements. However, three modifications are proposed to improve the variable rate compression performance. First, multi objective optimization is used for (post) training. Second, a quantization-reconstruction offset is introduced into the quantization operation. Third, variable rate quantization is used also for the hyper latent. All these modifications can be made on a pre-trained single-rate compression model by performing post training. The algorithms are implemented into three well-known image compression models and the achieved variable rate compression results indicate negligible or minimal compression performance loss compared to training multiple models. (Codes will be shared at https://github.com/InterDigitalInc/CompressAI)
While various deep learning methods were proposed for low-dose computed tomography (CT) denoising, they often suffer from over-smoothing, blurring, and lack of explainability. To alleviate these issues, we propose a plug-and-play Language-Engaged Dual-space Alignment loss (LEDA) to optimize low-dose CT denoising models. Our idea is to leverage large language models (LLMs) to align denoised CT and normal dose CT images in both the continuous perceptual space and discrete semantic space, which is the first LLM-based scheme for low-dose CT denoising. LEDA involves two steps: the first is to pretrain an LLM-guided CT autoencoder, which can encode a CT image into continuous high-level features and quantize them into a token space to produce semantic tokens derived from the LLM's vocabulary; and the second is to minimize the discrepancy between the denoised CT images and normal dose CT in terms of both encoded high-level features and quantized token embeddings derived by the LLM-guided CT autoencoder. Extensive experimental results on two public LDCT denoising datasets demonstrate that our LEDA can enhance existing denoising models in terms of quantitative metrics and qualitative evaluation, and also provide explainability through language-level image understanding. Source code is available at https://github.com/hao1635/LEDA.
In recent years, text-image joint pre-training techniques have shown promising results in various tasks. However, in Optical Character Recognition (OCR) tasks, aligning text instances with their corresponding text regions in images poses a challenge, as it requires effective alignment between text and OCR-Text (referring to the text in images as OCR-Text to distinguish from the text in natural language) rather than a holistic understanding of the overall image content. In this paper, we propose a new pre-training method called OCR-Text Destylization Modeling (ODM) that transfers diverse styles of text found in images to a uniform style based on the text prompt. With ODM, we achieve better alignment between text and OCR-Text and enable pre-trained models to adapt to the complex and diverse styles of scene text detection and spotting tasks. Additionally, we have designed a new labeling generation method specifically for ODM and combined it with our proposed Text-Controller module to address the challenge of annotation costs in OCR tasks, allowing a larger amount of unlabeled data to participate in pre-training. Extensive experiments on multiple public datasets demonstrate that our method significantly improves performance and outperforms current pre-training methods in scene text detection and spotting tasks. Code is available at {https://github.com/PriNing/ODM}.
We introduce StyleGaussian, a novel 3D style transfer technique that allows instant transfer of any image's style to a 3D scene at 10 frames per second (fps). Leveraging 3D Gaussian Splatting (3DGS), StyleGaussian achieves style transfer without compromising its real-time rendering ability and multi-view consistency. It achieves instant style transfer with three steps: embedding, transfer, and decoding. Initially, 2D VGG scene features are embedded into reconstructed 3D Gaussians. Next, the embedded features are transformed according to a reference style image. Finally, the transformed features are decoded into the stylized RGB. StyleGaussian has two novel designs. The first is an efficient feature rendering strategy that first renders low-dimensional features and then maps them into high-dimensional features while embedding VGG features. It cuts the memory consumption significantly and enables 3DGS to render the high-dimensional memory-intensive features. The second is a K-nearest-neighbor-based 3D CNN. Working as the decoder for the stylized features, it eliminates the 2D CNN operations that compromise strict multi-view consistency. Extensive experiments show that StyleGaussian achieves instant 3D stylization with superior stylization quality while preserving real-time rendering and strict multi-view consistency. Project page: https://kunhao-liu.github.io/StyleGaussian/
Given the remarkable achievements in image generation through diffusion models, the research community has shown increasing interest in extending these models to video generation. Recent diffusion models for video generation have predominantly utilized attention layers to extract temporal features. However, attention layers are limited by their memory consumption, which increases quadratically with the length of the sequence. This limitation presents significant challenges when attempting to generate longer video sequences using diffusion models. To overcome this challenge, we propose leveraging state-space models (SSMs). SSMs have recently gained attention as viable alternatives due to their linear memory consumption relative to sequence length. In the experiments, we first evaluate our SSM-based model with UCF101, a standard benchmark of video generation. In addition, to investigate the potential of SSMs for longer video generation, we perform an experiment using the MineRL Navigate dataset, varying the number of frames to 64 and 150. In these settings, our SSM-based model can considerably save memory consumption for longer sequences, while maintaining competitive FVD scores to the attention-based models. Our codes are available at https://github.com/shim0114/SSM-Meets-Video-Diffusion-Models.
In this paper, we explore a new generative approach for learning visual representations. Our method, DARL, employs a decoder-only Transformer to predict image patches autoregressively. We find that training with Mean Squared Error (MSE) alone leads to strong representations. To enhance the image generation ability, we replace the MSE loss with the diffusion objective by using a denoising patch decoder. We show that the learned representation can be improved by using tailored noise schedules and longer training in larger models. Notably, the optimal schedule differs significantly from the typical ones used in standard image diffusion models. Overall, despite its simple architecture, DARL delivers performance remarkably close to state-of-the-art masked prediction models under the fine-tuning protocol. This marks an important step towards a unified model capable of both visual perception and generation, effectively combining the strengths of autoregressive and denoising diffusion models.
Super-resolution (SR) is an ill-posed inverse problem, where the size of the set of feasible solutions that are consistent with a given low-resolution image is very large. Many algorithms have been proposed to find a "good" solution among the feasible solutions that strike a balance between fidelity and perceptual quality. Unfortunately, all known methods generate artifacts and hallucinations while trying to reconstruct high-frequency (HF) image details. A fundamental question is: Can a model learn to distinguish genuine image details from artifacts? Although some recent works focused on the differentiation of details and artifacts, this is a very challenging problem and a satisfactory solution is yet to be found. This paper shows that the characterization of genuine HF details versus artifacts can be better learned by training GAN-based SR models using wavelet-domain loss functions compared to RGB-domain or Fourier-space losses. Although wavelet-domain losses have been used in the literature before, they have not been used in the context of the SR task. More specifically, we train the discriminator only on the HF wavelet sub-bands instead of on RGB images and the generator is trained by a fidelity loss over wavelet subbands to make it sensitive to the scale and orientation of structures. Extensive experimental results demonstrate that our model achieves better perception-distortion trade-off according to multiple objective measures and visual evaluations.
Abstract reasoning problems challenge the perceptual and cognitive abilities of AI algorithms, demanding deeper pattern discernment and inductive reasoning beyond explicit image features. This study introduces PMoC, a tailored probability model for the Bongard-Logo problem, achieving high reasoning accuracy by constructing independent probability models. Additionally, we present Pose-Transformer, an enhanced Transformer-Encoder designed for complex abstract reasoning tasks, including Bongard-Logo, RAVEN, I-RAVEN, and PGM. Pose-Transformer incorporates positional information learning, inspired by capsule networks' pose matrices, enhancing its focus on local positional relationships in image data processing. When integrated with PMoC, it further improves reasoning accuracy. Our approach effectively addresses reasoning difficulties associated with abstract entities' positional changes, outperforming previous models on the OIG, D3$\times$3 subsets of RAVEN, and PGM databases. This research contributes to advancing AI's capabilities in abstract reasoning and cognitive pattern recognition.
In computational pathology, random sampling of patches during training of Multiple Instance Learning (MIL) methods is computationally efficient and serves as a regularization strategy. Despite its promising benefits, questions concerning performance trends for varying sample sizes and its influence on model interpretability remain. Addressing these, we reach an optimal performance enhancement of 1.7% using thirty percent of patches on the CAMELYON16 dataset, and 3.7% with only eight samples on the TUPAC16 dataset. We also find interpretability effects are strongly dataset-dependent, with interpretability impacted on CAMELYON16, while remaining unaffected on TUPAC16. This reinforces that both the performance and interpretability relationships with sampling are closely task-specific. End-to-end training with 1024 samples reveals improvements across both datasets compared to pre-extracted features, further highlighting the potential of this efficient approach.