Visual language tasks require AI models to comprehend and reason with both visual and textual content. Driven by the power of Large Language Models (LLMs), two prominent methods have emerged: (1) the hybrid integration between LLMs and Vision-Language Models (VLMs), where visual inputs are firstly converted into language descriptions by VLMs, serving as inputs for LLMs to generate final answer(s); (2) visual feature alignment in language space, where visual inputs are encoded as embeddings and projected to LLMs' language space via further supervised fine-tuning. The first approach provides light training costs and interpretability but is hard to be optimized in an end-to-end fashion. The second approach presents decent performance, but feature alignment usually requires large amounts of training data and lacks interpretability. To tackle this dilemma, we propose a novel approach, Inner Monologue Multi-Modal Optimization (IMMO), to solve complex vision language problems by simulating inner monologue processes, a cognitive process in which an individual engages in silent verbal communication with themselves. We enable LLMs and VLMs to interact through natural language conversation and propose to use a two-stage training process to learn how to do the inner monologue (self-asking questions and answering questions). IMMO is evaluated on two popular tasks and the results suggest by emulating the cognitive phenomenon of internal dialogue, our approach can enhance reasoning and explanation abilities, contributing to the more effective fusion of vision and language models. More importantly, instead of using predefined human-crafted monologues, IMMO learns this process within the deep learning models, promising wider applicability to many different AI problems beyond vision language tasks.
Machine learning models often learn to make predictions that rely on sensitive social attributes like gender and race, which poses significant fairness risks, especially in societal applications, such as hiring, banking, and criminal justice. Existing work tackles this issue by minimizing the employed information about social attributes in models for debiasing. However, the high correlation between target task and these social attributes makes learning on the target task incompatible with debiasing. Given that model bias arises due to the learning of bias features (\emph{i.e}., gender) that help target task optimization, we explore the following research question: \emph{Can we leverage shortcut features to replace the role of bias feature in target task optimization for debiasing?} To this end, we propose \emph{Shortcut Debiasing}, to first transfer the target task's learning of bias attributes from bias features to shortcut features, and then employ causal intervention to eliminate shortcut features during inference. The key idea of \emph{Shortcut Debiasing} is to design controllable shortcut features to on one hand replace bias features in contributing to the target task during the training stage, and on the other hand be easily removed by intervention during the inference stage. This guarantees the learning of the target task does not hinder the elimination of bias features. We apply \emph{Shortcut Debiasing} to several benchmark datasets, and achieve significant improvements over the state-of-the-art debiasing methods in both accuracy and fairness.
In this paper, we explore the zero-shot capability of the Segment Anything Model (SAM) for food image segmentation. To address the lack of class-specific information in SAM-generated masks, we propose a novel framework, called FoodSAM. This innovative approach integrates the coarse semantic mask with SAM-generated masks to enhance semantic segmentation quality. Besides, we recognize that the ingredients in food can be supposed as independent individuals, which motivated us to perform instance segmentation on food images. Furthermore, FoodSAM extends its zero-shot capability to encompass panoptic segmentation by incorporating an object detector, which renders FoodSAM to effectively capture non-food object information. Drawing inspiration from the recent success of promptable segmentation, we also extend FoodSAM to promptable segmentation, supporting various prompt variants. Consequently, FoodSAM emerges as an all-encompassing solution capable of segmenting food items at multiple levels of granularity. Remarkably, this pioneering framework stands as the first-ever work to achieve instance, panoptic, and promptable segmentation on food images. Extensive experiments demonstrate the feasibility and impressing performance of FoodSAM, validating SAM's potential as a prominent and influential tool within the domain of food image segmentation. We release our code at https://github.com/jamesjg/FoodSAM.
Spotting facial micro-expression from videos finds various potential applications in fields including clinical diagnosis and interrogation, meanwhile this task is still difficult due to the limited scale of training data. To solve this problem, this paper tries to formulate a new task called micro-expression generation and then presents a strong baseline which combines the first order motion model with facial prior knowledge. Given a target face, we intend to drive the face to generate micro-expression videos according to the motion patterns of source videos. Specifically, our new model involves three modules. First, we extract facial prior features from a region focusing module. Second, we estimate facial motion using key points and local affine transformations with a motion prediction module. Third, expression generation module is used to drive the target face to generate videos. We train our model on public CASME II, SAMM and SMIC datasets and then use the model to generate new micro-expression videos for evaluation. Our model achieves the first place in the Facial Micro-Expression Challenge 2021 (MEGC2021), where our superior performance is verified by three experts with Facial Action Coding System certification. Source code is provided in https://github.com/Necolizer/Facial-Prior-Based-FOMM.
Deep learning for unsupervised image segmentation remains challenging due to the absence of human labels. The common idea is to train a segmentation head, with the supervision of pixel-wise pseudo-labels generated based on the representation of self-supervised backbones. By doing so, the model performance depends much on the distance between the distributions of target datasets and the pre-training dataset (e.g., ImageNet). In this work, we investigate a new task, namely unsupervised camouflaged object segmentation (UCOS), where the target objects own a common rarely-seen attribute, i.e., camouflage. Unsurprisingly, we find that the state-of-the-art unsupervised models struggle in adapting UCOS, due to the domain gap between the properties of generic and camouflaged objects. To this end, we formulate the UCOS as a source-free unsupervised domain adaptation task (UCOS-DA), where both source labels and target labels are absent during the whole model training process. Specifically, we define a source model consisting of self-supervised vision transformers pre-trained on ImageNet. On the other hand, the target domain includes a simple linear layer (i.e., our target model) and unlabeled camouflaged objects. We then design a pipeline for foreground-background-contrastive self-adversarial domain adaptation, to achieve robust UCOS. As a result, our baseline model achieves superior segmentation performance when compared with competing unsupervised models on the UCOS benchmark, with the training set which's scale is only one tenth of the supervised COS counterpart.
Dialogue act annotations are important to improve response generation quality in task-oriented dialogue systems. However, it can be challenging to use dialogue acts to control response generation in a generalizable way because different datasets and tasks may have incompatible annotations. While alternative methods that utilize latent action spaces or reinforcement learning do not require explicit annotations, they may lack interpretability or face difficulties defining task-specific rewards. In this work, we present a novel end-to-end latent dialogue act model (DiactTOD) that represents dialogue acts in a latent space. DiactTOD, when pre-trained on a large corpus, is able to predict and control dialogue acts to generate controllable responses using these latent representations in a zero-shot fashion. Our approach demonstrates state-of-the-art performance across a wide range of experimental settings on the MultiWOZ dataset, including zero-shot, few-shot, and full data fine-tuning with both end-to-end and policy optimization configurations.
Palmprint as biometrics has gained increasing attention recently due to its discriminative ability and robustness. However, existing methods mainly improve palmprint verification within one spectrum, which is challenging to verify across different spectrums. Additionally, in distributed server-client-based deployment, palmprint verification systems predominantly necessitate clients to transmit private data for model training on the centralized server, thereby engendering privacy apprehensions. To alleviate the above issues, in this paper, we propose a physics-driven spectrum-consistent federated learning method for palmprint verification, dubbed as PSFed-Palm. PSFed-Palm draws upon the inherent physical properties of distinct wavelength spectrums, wherein images acquired under similar wavelengths display heightened resemblances. Our approach first partitions clients into short- and long-spectrum groups according to the wavelength range of their local spectrum images. Subsequently, we introduce anchor models for short- and long-spectrum, which constrain the optimization directions of local models associated with long- and short-spectrum images. Specifically, a spectrum-consistent loss that enforces the model parameters and feature representation to align with their corresponding anchor models is designed. Finally, we impose constraints on the local models to ensure their consistency with the global model, effectively preventing model drift. This measure guarantees spectrum consistency while protecting data privacy, as there is no need to share local data. Extensive experiments are conducted to validate the efficacy of our proposed PSFed-Palm approach. The proposed PSFed-Palm demonstrates compelling performance despite only a limited number of training data. The codes will be released at https://github.com/Zi-YuanYang/PSFed-Palm.
The whole slide image (WSI) classification is often formulated as a multiple instance learning (MIL) problem. Since the positive tissue is only a small fraction of the gigapixel WSI, existing MIL methods intuitively focus on identifying salient instances via attention mechanisms. However, this leads to a bias towards easy-to-classify instances while neglecting hard-to-classify instances. Some literature has revealed that hard examples are beneficial for modeling a discriminative boundary accurately. By applying such an idea at the instance level, we elaborate a novel MIL framework with masked hard instance mining (MHIM-MIL), which uses a Siamese structure (Teacher-Student) with a consistency constraint to explore the potential hard instances. With several instance masking strategies based on attention scores, MHIM-MIL employs a momentum teacher to implicitly mine hard instances for training the student model, which can be any attention-based MIL model. This counter-intuitive strategy essentially enables the student to learn a better discriminating boundary. Moreover, the student is used to update the teacher with an exponential moving average (EMA), which in turn identifies new hard instances for subsequent training iterations and stabilizes the optimization. Experimental results on the CAMELYON-16 and TCGA Lung Cancer datasets demonstrate that MHIM-MIL outperforms other latest methods in terms of performance and training cost. The code is available at: https://github.com/DearCaat/MHIM-MIL.
Large-scale pre-trained Vision-Language Models (VLMs), such as CLIP, establish the correlation between texts and images, achieving remarkable success on various downstream tasks with fine-tuning. In existing fine-tuning methods, the class-specific text description is matched against the whole image. We recognize that this whole image matching is not effective since images from the same class often contain a set of different semantic objects, and an object further consists of a set of semantic parts or concepts. Individual semantic parts or concepts may appear in image samples from different classes. To address this issue, in this paper, we develop a new method called cross-model concept learning and inference (CCLI). Using the powerful text-image correlation capability of CLIP, our method automatically learns a large set of distinctive visual concepts from images using a set of semantic text concepts. Based on these visual concepts, we construct a discriminative representation of images and learn a concept inference network to perform downstream image classification tasks, such as few-shot learning and domain generalization. Extensive experimental results demonstrate that our CCLI method is able to improve the performance upon the current state-of-the-art methods by large margins, for example, by up to 8.0% improvement on few-shot learning and by up to 1.3% for domain generalization.
While various deep learning methods have been proposed for low-dose computed tomography (CT) denoising, most of them leverage the normal-dose CT images as the ground-truth to supervise the denoising process. These methods typically ignore the inherent correlation within a single CT image, especially the anatomical semantics of human tissues, and lack the interpretability on the denoising process. In this paper, we propose a novel Anatomy-aware Supervised CONtrastive learning framework, termed ASCON, which can explore the anatomical semantics for low-dose CT denoising while providing anatomical interpretability. The proposed ASCON consists of two novel designs: an efficient self-attention-based U-Net (ESAU-Net) and a multi-scale anatomical contrastive network (MAC-Net). First, to better capture global-local interactions and adapt to the high-resolution input, an efficient ESAU-Net is introduced by using a channel-wise self-attention mechanism. Second, MAC-Net incorporates a patch-wise non-contrastive module to capture inherent anatomical information and a pixel-wise contrastive module to maintain intrinsic anatomical consistency. Extensive experimental results on two public low-dose CT denoising datasets demonstrate superior performance of ASCON over state-of-the-art models. Remarkably, our ASCON provides anatomical interpretability for low-dose CT denoising for the first time. Source code is available at https://github.com/hao1635/ASCON.