Abstract:Deep learning has provided considerable advancements for multimedia systems, yet the interpretability of deep models remains a challenge. State-of-the-art post-hoc explainability methods, such as GradCAM, provide visual interpretation based on heatmaps but lack conceptual clarity. Prototype-based approaches, like ProtoPNet and PIPNet, offer a more structured explanation but rely on fixed patches, limiting their robustness and semantic consistency. To address these limitations, a part-prototypical concept mining network (PCMNet) is proposed that dynamically learns interpretable prototypes from meaningful regions. PCMNet clusters prototypes into concept groups, creating semantically grounded explanations without requiring additional annotations. Through a joint process of unsupervised part discovery and concept activation vector extraction, PCMNet effectively captures discriminative concepts and makes interpretable classification decisions. Our extensive experiments comparing PCMNet against state-of-the-art methods on multiple datasets show that it can provide a high level of interpretability, stability, and robustness under clean and occluded scenarios.
Abstract:Weakly supervised object localization (WSOL) methods allow training models to classify images and localize ROIs. WSOL only requires low-cost image-class annotations yet provides a visually interpretable classifier, which is important in histology image analysis. Standard WSOL methods rely on class activation mapping (CAM) methods to produce spatial localization maps according to a single- or two-step strategy. While both strategies have made significant progress, they still face several limitations with histology images. Single-step methods can easily result in under- or over-activation due to the limited visual ROI saliency in histology images and the limited localization cues. They also face the well-known issue of asynchronous convergence between classification and localization tasks. The two-step approach is sub-optimal because it is tied to a frozen classifier, limiting the capacity for localization. Moreover, these methods also struggle when applied to out-of-distribution (OOD) datasets. In this paper, a multi-task approach for WSOL is introduced for simultaneous training of both tasks to address the asynchronous convergence problem. In particular, localization is performed in the pixel-feature space of an image encoder that is shared with classification. This allows learning discriminant features and accurate delineation of foreground/background regions to support ROI localization and image classification. We propose PixelCAM, a cost-effective foreground/background pixel-wise classifier in the pixel-feature space that allows for spatial object localization. PixelCAM is trained using pixel pseudo-labels collected from a pretrained WSOL model. Both image and pixel-wise classifiers are trained simultaneously using standard gradient descent. In addition, our pixel classifier can easily be integrated into CNN- and transformer-based architectures without any modifications.
Abstract:Various deep learning models have been developed for indoor localization based on radio-frequency identification (RFID) tags. However, they often require adaptation to ensure accurate tracking in new target operational domains. To address this challenge, unsupervised domain adaptation (UDA) methods have been proposed to align pre-trained models with data from target environments. However, they rely on large annotated datasets from the initial domain (source). Source data access is limited by privacy, storage, computational, and transfer constraints. Although many source-free domain adaptation (SFDA) methods address these constraints in classification, applying them to regression models for localization remains challenging. Indeed, target datasets for indoor localization are typically small, with few features and samples, and are noisy. Adapting regression models requires high-confidence target pseudo-annotation to avoid over-training. In this paper, a specialized mean-teacher method called MTLoc is proposed for SFDA. MTLoc updates the student network using noisy data and teacher-generated pseudo-labels. The teacher network maintains stability through exponential moving averages. To further ensure robustness, the teacher's pseudo-labels are refined using k-nearest neighbor correction. MTLoc allows for self-supervised learning on target data, facilitating effective adaptation to dynamic and noisy indoor environments. Validated using real-world data from our experimental setup with INLAN Inc., our results show that MTLoc achieves high localization accuracy under challenging conditions, significantly reducing localization error compared to baselines, including the state-of-the-art adversarial UDA approach with access to source data.
Abstract:Image retrieval methods rely on metric learning to train backbone feature extraction models that can extract discriminant queries and reference (gallery) feature representations for similarity matching. Although state-of-the-art accuracy has improved considerably with the advent of deep learning (DL) models trained on large datasets, image retrieval remains challenging in many real-world video analytics and surveillance applications, e.g., person re-identification. Using the Euclidean space for matching limits the performance in real-world applications due to the curse of dimensionality, overfitting, and sensitivity to noisy data. We argue that the feature dissimilarity space is more suitable for similarity matching, and propose a dichotomy transformation to project query and reference embeddings into a single embedding in the dissimilarity space. We also advocate for end-to-end training of a backbone and binary classification models for pair-wise matching. As opposed to comparing the distance between queries and reference embeddings, we show the benefits of classifying the single dissimilarity space embedding (as similar or dissimilar), especially when trained end-to-end. We propose a method to train the max-margin classifier together with the backbone feature extractor by applying constraints to the L2 norm of the classifier weights along with the hinge loss. Our extensive experiments on challenging image retrieval datasets and using diverse feature extraction backbones highlight the benefits of similarity matching in the dissimilarity space. In particular, when jointly training the feature extraction backbone and regularised classifier for matching, the dissimilarity space provides a higher level of accuracy.
Abstract:The zero-shot performance of object detectors degrades when tested on different modalities, such as infrared and depth. While recent work has explored image translation techniques to adapt detectors to new modalities, these methods are limited to a single modality and apply only to traditional detectors. Recently, vision-language detectors, such as YOLO-World and Grounding DINO, have shown promising zero-shot capabilities, however, they have not yet been adapted for other visual modalities. Traditional fine-tuning approaches tend to compromise the zero-shot capabilities of the detectors. The visual prompt strategies commonly used for classification with vision-language models apply the same linear prompt translation to each image making them less effective. To address these limitations, we propose ModPrompt, a visual prompt strategy to adapt vision-language detectors to new modalities without degrading zero-shot performance. In particular, an encoder-decoder visual prompt strategy is proposed, further enhanced by the integration of inference-friendly task residuals, facilitating more robust adaptation. Empirically, we benchmark our method for modality adaptation on two vision-language detectors, YOLO-World and Grounding DINO, and on challenging infrared (LLVIP, FLIR) and depth (NYUv2) data, achieving performance comparable to full fine-tuning while preserving the model's zero-shot capability. Our code is available at: https://github.com/heitorrapela/ModPrompt
Abstract:Diffusion models have emerged as highly effective techniques for inpainting, however, they remain constrained by slow sampling rates. While recent advances have enhanced generation quality, they have also increased sampling time, thereby limiting scalability in real-world applications. We investigate the generative sampling process of diffusion-based inpainting models and observe that these models make minimal use of the input condition during the initial sampling steps. As a result, the sampling trajectory deviates from the data manifold, requiring complex synchronization mechanisms to realign the generation process. To address this, we propose Time-aware Diffusion Paint (TD-Paint), a novel approach that adapts the diffusion process by modeling variable noise levels at the pixel level. This technique allows the model to efficiently use known pixel values from the start, guiding the generation process toward the target manifold. By embedding this information early in the diffusion process, TD-Paint significantly accelerates sampling without compromising image quality. Unlike conventional diffusion-based inpainting models, which require a dedicated architecture or an expensive generation loop, TD-Paint achieves faster sampling times without architectural modifications. Experimental results across three datasets show that TD-Paint outperforms state-of-the-art diffusion models while maintaining lower complexity.
Abstract:Source-free domain adaptation (SFDA) is a challenging problem in object detection, where a pre-trained source model is adapted to a new target domain without using any source domain data for privacy and efficiency reasons. Most state-of-the-art SFDA methods for object detection have been proposed for Faster-RCNN, a detector that is known to have high computational complexity. This paper focuses on domain adaptation techniques for real-world vision systems, particularly for the YOLO family of single-shot detectors known for their fast baselines and practical applications. Our proposed SFDA method - Source-Free YOLO (SF-YOLO) - relies on a teacher-student framework in which the student receives images with a learned, target domain-specific augmentation, allowing the model to be trained with only unlabeled target data and without requiring feature alignment. A challenge with self-training using a mean-teacher architecture in the absence of labels is the rapid decline of accuracy due to noisy or drifting pseudo-labels. To address this issue, a teacher-to-student communication mechanism is introduced to help stabilize the training and reduce the reliance on annotated target data for model selection. Despite its simplicity, our approach is competitive with state-of-the-art detectors on several challenging benchmark datasets, even sometimes outperforming methods that use source data for adaptation.
Abstract:Human emotion is a complex phenomenon conveyed and perceived through facial expressions, vocal tones, body language, and physiological signals. Multimodal emotion recognition systems can perform well because they can learn complementary and redundant semantic information from diverse sensors. In real-world scenarios, only a subset of the modalities employed for training may be available at test time. Learning privileged information allows a model to exploit data from additional modalities that are only available during training. SOTA methods for PKD have been proposed to distill information from a teacher model (with privileged modalities) to a student model (without privileged modalities). However, such PKD methods utilize point-to-point matching and do not explicitly capture the relational information. Recently, methods have been proposed to distill the structural information. However, PKD methods based on structural similarity are primarily confined to learning from a single joint teacher representation, which limits their robustness, accuracy, and ability to learn from diverse multimodal sources. In this paper, a multi-teacher PKD (MT-PKDOT) method with self-distillation is introduced to align diverse teacher representations before distilling them to the student. MT-PKDOT employs a structural similarity KD mechanism based on a regularized optimal transport (OT) for distillation. The proposed MT-PKDOT method was validated on the Affwild2 and Biovid datasets. Results indicate that our proposed method can outperform SOTA PKD methods. It improves the visual-only baseline on Biovid data by 5.5%. On the Affwild2 dataset, the proposed method improves 3% and 5% over the visual-only baseline for valence and arousal respectively. Allowing the student to learn from multiple diverse sources is shown to increase the accuracy and implicitly avoids negative transfer to the student model.
Abstract:Systems for multimodal Emotion Recognition (ER) commonly rely on features extracted from different modalities (e.g., visual, audio, and textual) to predict the seven basic emotions. However, compound emotions often occur in real-world scenarios and are more difficult to predict. Compound multimodal ER becomes more challenging in videos due to the added uncertainty of diverse modalities. In addition, standard features-based models may not fully capture the complex and subtle cues needed to understand compound emotions. %%%% Since relevant cues can be extracted in the form of text, we advocate for textualizing all modalities, such as visual and audio, to harness the capacity of large language models (LLMs). These models may understand the complex interaction between modalities and the subtleties of complex emotions. Although training an LLM requires large-scale datasets, a recent surge of pre-trained LLMs, such as BERT and LLaMA, can be easily fine-tuned for downstream tasks like compound ER. This paper compares two multimodal modeling approaches for compound ER in videos -- standard feature-based vs. text-based. Experiments were conducted on the challenging C-EXPR-DB dataset for compound ER, and contrasted with results on the MELD dataset for basic ER. Our code is available
Abstract:Weakly-Supervised Video Object Localization (WSVOL) involves localizing an object in videos using only video-level labels, also referred to as tags. State-of-the-art WSVOL methods like Temporal CAM (TCAM) rely on class activation mapping (CAM) and typically require a pre-trained CNN classifier. However, their localization accuracy is affected by their tendency to minimize the mutual information between different instances of a class and exploit temporal information during training for downstream tasks, e.g., detection and tracking. In the absence of bounding box annotation, it is challenging to exploit precise information about objects from temporal cues because the model struggles to locate objects over time. To address these issues, a novel method called transformer based CAM for videos (TrCAM-V), is proposed for WSVOL. It consists of a DeiT backbone with two heads for classification and localization. The classification head is trained using standard classification loss (CL), while the localization head is trained using pseudo-labels that are extracted using a pre-trained CLIP model. From these pseudo-labels, the high and low activation values are considered to be foreground and background regions, respectively. Our TrCAM-V method allows training a localization network by sampling pseudo-pixels on the fly from these regions. Additionally, a conditional random field (CRF) loss is employed to align the object boundaries with the foreground map. During inference, the model can process individual frames for real-time localization applications. Extensive experiments on challenging YouTube-Objects unconstrained video datasets show that our TrCAM-V method achieves new state-of-the-art performance in terms of classification and localization accuracy.