Medical image synthesis presents unique challenges due to the inherent complexity and high-resolution details required in clinical contexts. Traditional generative architectures such as Generative Adversarial Networks (GANs) or Variational Auto Encoder (VAEs) have shown great promise for high-resolution image generation but struggle with preserving fine-grained details that are key for accurate diagnosis. To address this issue, we introduce Pixel Perfect MegaMed, the first vision-language foundation model to synthesize images at resolutions of 1024x1024. Our method deploys a multi-scale transformer architecture designed specifically for ultra-high resolution medical image generation, enabling the preservation of both global anatomical context and local image-level details. By leveraging vision-language alignment techniques tailored to medical terminology and imaging modalities, Pixel Perfect MegaMed bridges the gap between textual descriptions and visual representations at unprecedented resolution levels. We apply our model to the CheXpert dataset and demonstrate its ability to generate clinically faithful chest X-rays from text prompts. Beyond visual quality, these high-resolution synthetic images prove valuable for downstream tasks such as classification, showing measurable performance gains when used for data augmentation, particularly in low-data regimes. Our code is accessible through the project website - https://tehraninasab.github.io/pixelperfect-megamed.
3D 객체의 방향을 정규화(canonicalization)하는 것은 3D 이해 및 분석에 필수적입니다. 기존 방식들은 종종 기하학적 단서에 의존하지만, 3D 정규화는 궁극적으로 의미론적으로 유의미한 방향을 요구합니다. 이러한 간극을 해결하기 위해, 본 연구에서는 정규화 전용 학습(training)이나 범주별 템플릿 없이도 동결된(frozen) 이미지-투-3D 생성 모델(generative model)의 의미론적 방향 사전 확률(prior)을 3D 정규화에 도입하는 범주 독립적인 생성 기반 프레임워크인 CANIS를 제안합니다. 구체적으로, CANIS는 먼저 후보 시점에서 입력 객체를 렌더링하고 유익한 시점을 선택하여 정규화된 방향의 프록시(proxy)를 생성합니다. 생성 과정에서 입력으로부터 인코딩된 희소 구조 잠재 변수(sparse structural latent)는 프록시가 객체의 기하학적 구조를 보존하도록 유도합니다. 그런 다음 CANIS는 선택된 이미지를 입력과 프록시 사이의 의미론적 브리지(semantic bridge)로 활용합니다. 이미지 패치는 프록시의 의미론적 영역을 식별하고, 깊이 역투영(depth back-projection)을 통해 입력에서 해당 영역을 찾아냅니다. 그 결과 생성된 의미론적 앵커(semantic anchor)는 기하학적 매칭을 제약하며, 이를 바탕으로 입력을 정규화하는 강체 변환(rigid transformation)을 추정합니다. 합성 벤치마크(benchmark)에 대한 실험을 통해 CANIS와 그 핵심 구성 요소를 검증하였으며, 부분 관측 데이터 및 OmniObject3D에 대한 정성적 결과는 불완전한 실제 스캔 데이터에 대한 적용 가능성을 보여줍니다. 또한 CANIS는 임의의 회전 상황에서 다운스트림 3D 분류(classification), 파트 분할(part segmentation), 조밀한 대응점 매칭(dense correspondence) 성능을 향상시킵니다. 프로젝트 페이지: https://kenkenzaii.github.io/Canis.
The purpose of mid-level visual element discovery is to find clusters of image patches that are both representative and discriminative. Here we study this problem from the prospective of pattern mining while relying on the recently popularized Convolutional Neural Networks (CNNs). We observe that a fully-connected CNN activation extracted from an image patch typically possesses two appealing properties that enable its seamless integration with pattern mining techniques. The marriage between CNN activations and association rule mining, a well-known pattern mining technique in the literature, leads to fast and effective discovery of representative and discriminative patterns from a huge number of image patches. When we retrieve and visualize image patches with the same pattern, surprisingly, they are not only visually similar but also semantically consistent, and thus give rise to a mid-level visual element in our work. Given the patterns and retrieved mid-level visual elements, we propose two methods to generate image feature representations for each. The first method is to use the patterns as codewords in a dictionary, similar to the Bag-of-Visual-Words model, we compute a Bag-of-Patterns representation. The second one relies on the retrieved mid-level visual elements to construct a Bag-of-Elements representation. We evaluate the two encoding methods on scene and object classification tasks, and demonstrate that our approach outperforms or matches recent works using CNN activations for these tasks.
Zero-shot sketch-based image retrieval (ZS-SBIR) is a task of cross-domain image retrieval from a natural image gallery with free-hand sketch under a zero-shot scenario. Previous works mostly focus on a generative approach that takes a highly abstract and sparse sketch as input and then synthesizes the corresponding natural image. However, the intrinsic visual sparsity and large intra-class variance of the sketch make the learning of the conditional decoder more difficult and hence achieve unsatisfactory retrieval performance. In this paper, we propose a novel stacked semantic-guided network to address the unique characteristics of sketches in ZS-SBIR. Specifically, we devise multi-layer feature fusion networks that incorporate different intermediate feature representation information in a deep neural network to alleviate the intrinsic sparsity of sketches. In order to improve visual knowledge transfer from seen to unseen classes, we elaborate a coarse-to-fine conditional decoder that generates coarse-grained category-specific corresponding features first (taking auxiliary semantic information as conditional input) and then generates fine-grained instance-specific corresponding features (taking sketch representation as conditional input). Furthermore, regression loss and classification loss are utilized to preserve the semantic and discriminative information of the synthesized features respectively. Extensive experiments on the large-scale Sketchy dataset and TU-Berlin dataset demonstrate that our proposed approach outperforms state-of-the-art methods by more than 20\% in retrieval performance.
Large-scale pre-training tasks like image classification, captioning, or self-supervised techniques do not incentivize learning the semantic boundaries of objects. However, recent generative foundation models built using text-based latent diffusion techniques may learn semantic boundaries. This is because they have to synthesize intricate details about all objects in an image based on a text description. Therefore, we present a technique for segmenting real and AI-generated images using latent diffusion models (LDMs) trained on internet-scale datasets. First, we show that the latent space of LDMs (z-space) is a better input representation compared to other feature representations like RGB images or CLIP encodings for text-based image segmentation. By training the segmentation models on the latent z-space, which creates a compressed representation across several domains like different forms of art, cartoons, illustrations, and photographs, we are also able to bridge the domain gap between real and AI-generated images. We show that the internal features of LDMs contain rich semantic information and present a technique in the form of LD-ZNet to further boost the performance of text-based segmentation. Overall, we show up to 6% improvement over standard baselines for text-to-image segmentation on natural images. For AI-generated imagery, we show close to 20% improvement compared to state-of-the-art techniques. The project is available at https://koutilya-pnvr.github.io/LD-ZNet/.
We propose a novel generative model architecture designed to learn representations for images that factor out a single attribute from the rest of the representation. A single object may have many attributes which when altered do not change the identity of the object itself. Consider the human face; the identity of a particular person is independent of whether or not they happen to be wearing glasses. The attribute of wearing glasses can be changed without changing the identity of the person. However, the ability to manipulate and alter image attributes without altering the object identity is not a trivial task. Here, we are interested in learning a representation of the image that separates the identity of an object (such as a human face) from an attribute (such as 'wearing glasses'). We demonstrate the success of our factorization approach by using the learned representation to synthesize the same face with and without a chosen attribute. We refer to this specific synthesis process as image attribute manipulation. We further demonstrate that our model achieves competitive scores, with state of the art, on a facial attribute classification task.
We compare the robustness of humans and current convolutional deep neural networks (DNNs) on object recognition under twelve different types of image degradations. First, using three well known DNNs (ResNet-152, VGG-19, GoogLeNet) we find the human visual system to be more robust to nearly all of the tested image manipulations, and we observe progressively diverging classification error-patterns between humans and DNNs when the signal gets weaker. Secondly, we show that DNNs trained directly on distorted images consistently surpass human performance on the exact distortion types they were trained on, yet they display extremely poor generalisation abilities when tested on other distortion types. For example, training on salt-and-pepper noise does not imply robustness on uniform white noise and vice versa. Thus, changes in the noise distribution between training and testing constitutes a crucial challenge to deep learning vision systems that can be systematically addressed in a lifelong machine learning approach. Our new dataset consisting of 83K carefully measured human psychophysical trials provide a useful reference for lifelong robustness against image degradations set by the human visual system.
Compressive sensing is a new technology for modern computational imaging systems. In comparison to widespread conventional image sensing, the compressive imaging paradigm requires specific forensic analysis techniques and tools. In this regards, one of basic scenarios in image forensics is to distinguish traditionally sensed images from sophisticated compressively sensed ones. To do this, we first mathematically and systematically model the imaging system based on compressive sensing technology. Afterwards, a simplified version of the whole model is presented, which is appropriate for forensic investigation applications. We estimate the nonlinear system of compressive sensing with a linear model. Then, we model the imaging pipeline as an inverse problem and demonstrate that different imagers have discriminative degradation kernels. Hence, blur kernels of various imaging systems have utilized as footprints for discriminating image acquisition sources. In order to accomplish the identification cycle, we have utilized the state-of-the-art Convolutional Neural Network (CNN) and Support Vector Machine (SVM) approaches to learn a classification system from estimated blur kernels. Numerical experiments show promising identification results. Simulation codes are available for research and development purposes.
Sum-Product Networks (SPNs) are hierarchical, graphical models that combine benefits of deep learning and probabilistic modeling. SPNs offer unique advantages to applications demanding exact probabilistic inference over high-dimensional, noisy inputs. Yet, compared to convolutional neural nets, they struggle with capturing complex spatial relationships in image data. To alleviate this issue, we introduce Deep Generalized Convolutional Sum-Product Networks (DGC-SPNs), which encode spatial features in a way similar to CNNs, while preserving the validity of the probabilistic SPN model. As opposed to existing SPN-based image representations, DGC-SPNs allow for overlapping convolution patches through a novel parameterization of dilations and strides, resulting in significantly improved feature coverage and feature resolution. DGC-SPNs substantially outperform other SPN architectures across several visual datasets and for both generative and discriminative tasks, including image inpainting and classification. These contributions are reinforced by the first simple, scalable, and GPU-optimized implementation of SPNs, integrated with the widely used Keras/TensorFlow framework. The resulting model is fully probabilistic and versatile, yet efficient and straightforward to apply in practical applications in place of traditional deep nets.
Identifying and characterizing the patient's blood samples is indispensable in diagnostics of malignance suspicious. A painstaking and sometimes subjective task is used in laboratories to manually classify white blood cells. Neural mathematical methods as deep learnings can be very useful in the automated recognition of blood cells. This study uses a particular type of deep learning i.e., convolutional neural networks (CNNs or ConvNets) for image recognition of the four (4) blood cell types (neutrophil, eosinophil, lymphocyte and monocyte) and to enable it to tag them employing a dataset of blood cells with labels for the corresponding cell types. The elements of the database are the input of our CNN and they allowed us to create learning models for the image recognition/classification of the blood cells. We evaluated the recognition performance and outputs learned by the networks in order to implement a neural image recognition model capable of distinguishing polynuclear cells (neutrophil and eosinophil) from those of mononuclear cells (lymphocyte and monocyte). The validation accuracy is 97.77%.
The importance of explainability in machine learning continues to grow, as both neural-network architectures and the data they model become increasingly complex. Unique challenges arise when a model's input features become high dimensional: on one hand, principled model-agnostic approaches to explainability become too computationally expensive; on the other, more efficient explainability algorithms lack natural interpretations for general users. In this work, we introduce a framework for human-interpretable explainability on high-dimensional data, consisting of two modules. First, we apply a semantically meaningful latent representation, both to reduce the raw dimensionality of the data, and to ensure its human interpretability. These latent features can be learnt, e.g. explicitly as disentangled representations or implicitly through image-to-image translation, or they can be based on any computable quantities the user chooses. Second, we adapt the Shapley paradigm for model-agnostic explainability to operate on these latent features. This leads to interpretable model explanations that are both theoretically controlled and computationally tractable. We benchmark our approach on synthetic data and demonstrate its effectiveness on several image-classification tasks.
Natural image matting separates the foreground from background in fractional occupancy which can be caused by highly transparent objects, complex foreground (e.g., net or tree), and/or objects containing very fine details (e.g., hairs). Although conventional matting formulation can be applied to all of the above cases, no previous work has attempted to reason the underlying causes of matting due to various foreground semantics. We show how to obtain better alpha mattes by incorporating into our framework semantic classification of matting regions. Specifically, we consider and learn 20 classes of matting patterns, and propose to extend the conventional trimap to semantic trimap. The proposed semantic trimap can be obtained automatically through patch structure analysis within trimap regions. Meanwhile, we learn a multi-class discriminator to regularize the alpha prediction at semantic level, and content-sensitive weights to balance different regularization losses. Experiments on multiple benchmarks show that our method outperforms other methods and has achieved the most competitive state-of-the-art performance. Finally, we contribute a large-scale Semantic Image Matting Dataset with careful consideration of data balancing across different semantic classes.
Contrastive learning is a powerful technique to learn representations that are semantically distinctive and geometrically invariant. While most of the earlier approaches have demonstrated its effectiveness on single-modality learning tasks such as image classification, recently there have been a few attempts towards extending this idea to multi-modal data. In this paper, we propose two loss functions based on normalized cross-entropy to perform the task of learning joint visual-semantic embedding using batch contrastive training. In a batch, for a given anchor point from one modality, we consider its negatives only from another modality, and define our first contrastive loss based on expected violations incurred by all the negatives. Next, we update this loss and define the second contrastive loss based on the violation incurred only by the hardest negative. We compare our results with existing visual-semantic embedding methods on cross-modal image-to-text and text-to-image retrieval tasks using the MS-COCO and Flickr30K datasets, where we outperform the state-of-the-art on the MS-COCO dataset and achieve comparable results on the Flickr30K dataset.
Vision transformers (ViTs) that model an image as a sequence of partitioned patches have shown notable performance in diverse vision tasks. Because partitioning patches eliminates the image structure, to reflect the order of patches, ViTs utilize an explicit component called positional embedding. However, we claim that the use of positional embedding does not simply guarantee the order-awareness of ViT. To support this claim, we analyze the actual behavior of ViTs using an effective receptive field. We demonstrate that during training, ViT acquires an understanding of patch order from the positional embedding that is trained to be a specific pattern. Based on this observation, we propose explicitly adding a Gaussian attention bias that guides the positional embedding to have the corresponding pattern from the beginning of training. We evaluated the influence of Gaussian attention bias on the performance of ViTs in several image classification, object detection, and semantic segmentation experiments. The results showed that proposed method not only facilitates ViTs to understand images but also boosts their performance on various datasets, including ImageNet, COCO 2017, and ADE20K.
Deep networks are frequently tuned to novel tasks and continue learning from ongoing data streams. Such sequential training requires consolidation of new and past information, a challenge predominantly addressed by retaining the most important data points - formally known as coresets. Traditionally, these coresets consist of entire samples, such as images or sentences. However, recent transformer architectures operate on tokens, leading to the famous assertion that an image is worth 16x16 words. Intuitively, not all of these tokens are equally informative or memorable. Going beyond coresets, we thus propose to construct a deeper-level data summary on the level of tokens. Our respectively named core tokensets both select the most informative data points and leverage feature attribution to store only their most relevant features. We demonstrate that core tokensets yield significant performance retention in incremental image classification, open-ended visual question answering, and continual image captioning with significantly reduced memory. In fact, we empirically find that a core tokenset of 1\% of the data performs comparably to at least a twice as large and up to 10 times larger coreset.
동물플랑크톤은 해양 먹이사슬의 1차 소비자로서 해양 생태 균형을 유지하는 데 중요한 역할을 한다. 그러나 Segment Anything Model(SAM)은 동물플랑크톤 특화 도메인 지식이 부족하여 현미경 이미지 인스턴스 분할(instance segmentation)에서 제한적인 성능을 보인다. 이러한 문제를 해결하기 위해, 본 논문에서는 부정확한 분류, 가늘고 긴 부속지(appendage)의 불연속적 분할, 불완전한 경계 분할과 같은 문제를 효과적으로 해결하는 SAM 및 웨이블릿 변환(wavelet transform) 기반의 새로운 인스턴스 분할 모델인 ZMIS-SAM을 제안한다. 제안하는 프레임워크는 세 가지 핵심 혁신을 포함한다. 먼저 ZM-ViT는 두 개의 경량 어댑터를 통해 동물플랑크톤 형태와 이미지 강도 분포를 모델링하는 SAM의 능력을 향상시킨다. 또한 인접 특징 집계 모듈(NFAM)은 범용 특징과 도메인 특화 특징을 통합하여 반투명하고 가늘고 긴 부속지의 연속적인 분할을 개선하며, 웨이블릿 기반 다중 스케일 다중 방향 특징 강화(WM2FE) 모듈은 고주파 세부 정보를 효과적으로 복원하여 경계 분할의 완성도를 정밀화한다. 광범위한 실험을 통해 ZMIS-SAM이 동물플랑크톤 데이터셋에서 최고 수준의 인스턴스 분할 성능을 달성하며, 여러 공개 교차 도메인 데이터셋 전반에서도 강력한 일반화 능력을 보임을 입증한다.
Although CNNs are widely considered as the state-of-the-art models in various applications of image analysis, one of the main challenges still open is the training of a CNN on high resolution images. Different strategies have been proposed involving either a rescaling of the image or an individual processing of parts of the image. Such strategies cannot be applied to images, such as gigapixel histopathological images, for which a high reduction in resolution inherently effects a loss of discriminative information, and in respect of which the analysis of single parts of the image suffers from a lack of global information or implies a high workload in terms of annotating the training images in such a way as to select significant parts. We propose a method for the analysis of gigapixel histopathological images solely by using weak image-level labels. In particular, two analysis tasks are taken into account: a binary classification and a prediction of the tumor proliferation score. Our method is based on a CNN structure consisting of a compressing path and a learning path. In the compressing path, the gigapixel image is packed into a grid-based feature map by using a residual network devoted to the feature extraction of each patch into which the image has been divided. In the learning path, attention modules are applied to the grid-based feature map, taking into account spatial correlations of neighboring patch features to find regions of interest, which are then used for the final whole slide analysis. Our method integrates both global and local information, is flexible with regard to the size of the input images and only requires weak image-level labels. Comparisons with different methods of the state-of-the-art on two well known datasets, Camelyon16 and TUPAC16, have been made to confirm the validity of the proposed model.
For many applications, utilizing DNNs (Deep Neural Networks) requires their implementation on a target architecture in an optimized manner concerning energy consumption, memory requirement, throughput, etc. DNN compression is used to reduce the memory footprint and complexity of a DNN before its deployment on hardware. Recent efforts to understand and explain AI (Artificial Intelligence) methods have led to a new research area, termed as explainable AI. Explainable AI methods allow us to understand better the inner working of DNNs, such as the importance of different neurons and features. The concepts from explainable AI provide an opportunity to improve DNN compression methods such as quantization and pruning in several ways that have not been sufficiently explored so far. In this paper, we utilize explainable AI methods: mainly DeepLIFT method. We use these methods for (1) pruning of DNNs; this includes structured and unstructured pruning of \ac{CNN} filters pruning as well as pruning weights of fully connected layers, (2) non-uniform quantization of DNN weights using clustering algorithm; this is also referred to as Weight Sharing, and (3) integer-based mixed-precision quantization; this is where each layer of a DNN may use a different number of integer bits. We use typical image classification datasets with common deep learning image classification models for evaluation. In all these three cases, we demonstrate significant improvements as well as new insights and opportunities from the use of explainable AI in DNN compression.
Recently, a massive number of deep learning based approaches have been successfully applied to various remote sensing image (RSI) recognition tasks. However, most existing advances of deep learning methods in the RSI field heavily rely on the features extracted by the manually designed backbone network, which severely hinders the potential of deep learning models due the complexity of RSI and the limitation of prior knowledge. In this paper, we research a new design paradigm for the backbone architecture in RSI recognition tasks, including scene classification, land-cover classification and object detection. A novel one-shot architecture search framework based on weight-sharing strategy and evolutionary algorithm is proposed, called RSBNet, which consists of three stages: Firstly, a supernet constructed in a layer-wise search space is pretrained on a self-assembled large-scale RSI dataset based on an ensemble single-path training strategy. Next, the pre-trained supernet is equipped with different recognition heads through the switchable recognition module and respectively fine-tuned on the target dataset to obtain task-specific supernet. Finally, we search the optimal backbone architecture for different recognition tasks based on the evolutionary algorithm without any network training. Extensive experiments have been conducted on five benchmark datasets for different recognition tasks, the results show the effectiveness of the proposed search paradigm and demonstrate that the searched backbone is able to flexibly adapt different RSI recognition tasks and achieve impressive performance.
Multi-label image classification datasets are often partially labeled where many labels are missing, posing a significant challenge to training accurate deep classifiers. However, the powerful Mixup sample-mixing data augmentation cannot be well utilized to address this challenge, as it cannot perform linear interpolation on the unknown labels to construct augmented samples. In this paper, we propose LogicMix, a Mixup variant designed for such partially labeled datasets. LogicMix mixes the sample labels by logical OR so that the unknown labels can be correctly mixed by utilizing OR's logical equivalences, including the domination and identity laws. Unlike Mixup, which mixes exactly two samples, LogicMix can mix multiple ($\geq2$) partially labeled samples, constructing visually more confused augmented samples to regularize training. LogicMix is more general and effective than other compared Mixup variants in the experiments on various partially labeled dataset scenarios. Moreover, it is plug-and-play and only requires minimal computation, hence it can be easily inserted into existing frameworks to collaborate with other methods to improve model performance with a negligible impact on training time, as demonstrated through extensive experiments. In particular, through the collaboration of LogicMix, RandAugment, Curriculum Labeling, and Category-wise Fine-Tuning, we attain state-of-the-art performance on MS-COCO, VG-200, and Pascal VOC 2007 benchmarking datasets. The remarkable generality, effectiveness, collaboration, and simplicity suggest that LogicMix promises to be a popular and vital data augmentation method.