The proceedings contain 698 papers. The topics discussed include: learning unbiased classifiers from biased data with meta-learning;robustness against gradient based attacks through cost effective network fine-tuning;...
ISBN:
(纸本)9798350302493
The proceedings contain 698 papers. The topics discussed include: learning unbiased classifiers from biased data with meta-learning;robustness against gradient based attacks through cost effective network fine-tuning;gradient attention balance network: mitigating face recognition racial bias via gradient attention;estimating and maximizing mutual information for knowledge distillation;synthetic sample selection for generalized zero-shot learning;training strategies for vision transformers for object detection;does image anonymization impact computervision training?;ultra-sonic sensor based object detection for autonomous vehicles;improvements to image reconstruction-based performance prediction for semantic segmentation in highly automated driving;zero-shot classification at different levels of granularity;difficulty estimation with action scores for computervision tasks;detail-preserving self-supervised monocular depth with self-supervised structural sharpening;isolated sign language recognition based on tree structure skeleton images;deep prototypical-parts ease morphological kidney stone identification and are competitively robust to photometric perturbations;wildlife image generation from scene graphs;towards characterizing the semantic robustness of face recognition;high-level context representation for emotion recognition in images;and mitigating catastrophic interference using unsupervised multi-part attention for RGB-IR face recognition.
The proceedings contain 802 papers. The topics discussed include: X-VARS: introducing explainability in football refereeing with multi-modal large language models;a hybrid ANN-SNN architecture for low-power and low-la...
ISBN:
(纸本)9798350365474
The proceedings contain 802 papers. The topics discussed include: X-VARS: introducing explainability in football refereeing with multi-modal large language models;a hybrid ANN-SNN architecture for low-power and low-latency visual perception;pseudo-label based unsupervised fine-tuning of a monocular 3D pose estimation model for sports motions;towards efficient audio-visual learners via empowering pre-trained vision transformers with cross-modal adaptation;a dual-mode approach for vision-based navigation in a lunar landing scenario;class similarity transition: decoupling class similarities and imbalance from generalized few-shot segmentation;ReweightOOD: loss reweighting for distance-based OOD detection;Hinge-Wasserstein: estimating multimodal aleatoric uncertainty in regression tasks;and ConPro: learning severity representation for medical images using contrastive learning and preference optimization.
The proceedings contain 355 papers. The topics discussed include: MultiNet++: multi-stream feature aggregation and geometric loss strategy for multi-task learning;privacy-preserving action recognition using coded aper...
ISBN:
(纸本)9781728125060
The proceedings contain 355 papers. The topics discussed include: MultiNet++: multi-stream feature aggregation and geometric loss strategy for multi-task learning;privacy-preserving action recognition using coded aperture videos;evading face recognition via partial tampering of faces;privacy-preserving annotation of face images through attribute-preserving face synthesis;towards deep neural network training on encrypted data;fooling automated surveillance cameras: adversarial patches to attack person detection;anonymousnet: natural face de-identification with measurable privacy;regularizer to mitigate gradient masking effect during single-step adversarial training;privacy preserving group membership verification and identification;defending against adversarial attacks using random forest;intersection to overpass: instance segmentation on filamentous structures with an orientation-aware neural network and terminus pairing algorithm;and surface parameterization and registration for statistical multiscale atlasing of organ development.
The proceedings contain 516 papers. The topics discussed include: OmniLayout: room layout reconstruction from indoor spherical panoramas;boosting adversarial robustness using feature level stochastic smoothing;beyond ...
ISBN:
(纸本)9781665448994
The proceedings contain 516 papers. The topics discussed include: OmniLayout: room layout reconstruction from indoor spherical panoramas;boosting adversarial robustness using feature level stochastic smoothing;beyond joint demosaicking and denoising: an image processing pipeline for a pixel-bin image sensor;assessment of deep learning based blood pressure prediction from PPG and rPPG signals;towards domain-specific explainable AI: model interpretation of a skin image classifier using a human approach;DAMSL: domain agnostic meta score-based learning;deep learning based spatial-temporal in-loop filtering for versatile video coding;automated tackle injury risk assessment in contact-based sports - a rugby union example;two-stage network for single image super-resolution;and ***: dataset for automatic mapping of buildings, woodlands, water and roads from aerial imagery.
The proceedings contain 523 papers. The topics discussed include: latent fingerprint image enhancement based on progressive generative adversarial network;zero-shot learning in the presence of hierarchically coarsened...
ISBN:
(纸本)9781728193601
The proceedings contain 523 papers. The topics discussed include: latent fingerprint image enhancement based on progressive generative adversarial network;zero-shot learning in the presence of hierarchically coarsened labels;multivariate confidence calibration for object detection;context-guided super-class inference for zero-shot detection;learning sparse ternary neural networks with entropy-constrained trained ternarization (EC2T);now that i can see, i can improve: enabling data-driven finetuning of CNNs on the edge;enhancing facial data diversity with style-based face aging;a simplified framework for zero-shot cross-modal sketch data retrieval;unsupervised single image super-resolution network (USISResNet) for real-world data using generative adversarial network;cross-regional oil palm tree detection;and leaf spot attention network for apple leaf disease identification.
The proceedings contain 561 papers. The topics discussed include: CORE: consistent representation learning for face forgery detection;aria: adversarially robust image attribution for content provenance;the reliability...
ISBN:
(纸本)9781665487399
The proceedings contain 561 papers. The topics discussed include: CORE: consistent representation learning for face forgery detection;aria: adversarially robust image attribution for content provenance;the reliability of forensic body-shape identification;detecting real-time deep-fake videos using active illumination;on the exploitation of deepfake model recognition;is synthetic voice detection research going into the right direction?;on improving cross-dataset generalization of deepfake detectors;rethinking adversarial examples in wargames;privacy leakage of adversarial training models in federated learning systems;towards comprehensive testing on the robustness of cooperative multi-agent reinforcement learning;robustness and adaptation to hidden factors of variation;adversarial robustness through the lens of convolutional filters;RODD: a self-supervised approach for robust out-of-distribution detection;an empirical study of data-free quantization’s tuning robustness;exploring robustness connection between artificial and natural adversarial examples;and adversarial machine learning attacks against video anomaly detection systems.
The proceedings contain 2356 papers. The topics discussed include: exploring discontinuity for video frame interpolation;two-view geometry scoring without correspondences;language-guided audio-visual source separation...
ISBN:
(纸本)9798350301298
The proceedings contain 2356 papers. The topics discussed include: exploring discontinuity for video frame interpolation;two-view geometry scoring without correspondences;language-guided audio-visual source separation via trimodal consistency;handwritten text generation from visual archetypes;Bayesian posterior approximation with stochastic ensembles;ERM-KTP: knowledge-level machine unlearning via knowledge transfer;PlenVDB: memory efficient VDB-based radiance fields for fast training and rendering;learning and aggregating lane graphs for urban automated driving;teaching matters: investigating the role of supervision in vision transformers;NeuralField-LDM: scene generation with hierarchical latent diffusion models;cut and learn for unsupervised object detection and instance segmentation;probabilistic debiasing of scene graphs;and unifying layout generation with a decoupled diffusion model.
Image anonymization is widely adapted in practice to comply with privacy regulations in many regions. However, anonymization often degrades the quality of the data, reducing its utility for computervision development...
详细信息
ISBN:
(纸本)9798350302493
Image anonymization is widely adapted in practice to comply with privacy regulations in many regions. However, anonymization often degrades the quality of the data, reducing its utility for computervision development. In this paper, we investigate the impact of image anonymization for training computervision models on key computervision tasks (detection, instance segmentation, and pose estimation). Specifically, we benchmark the recognition drop on common detection datasets, where we evaluate both traditional and realistic anonymization for faces and full bodies. Our comprehensive experiments reflect that traditional image anonymization substantially impacts final model performance, particularly when anonymizing the full body. Furthermore, we find that realistic anonymization can mitigate this decrease in performance, where our experiments reflect a minimal performance drop for face anonymization. Our study demonstrates that realistic anonymization can enable privacy-preserving computervision development with minimal performance degradation across a range of important computervision benchmarks.
Image completion is widely used in photo restoration and editing applications, e.g. for object removal. Recently, there has been a surge of research on generating diverse completions for missing regions. However, exis...
详细信息
ISBN:
(纸本)9798350302493
Image completion is widely used in photo restoration and editing applications, e.g. for object removal. Recently, there has been a surge of research on generating diverse completions for missing regions. However, existing methods require large training sets from a specific domain of interest, and often fail on general-content images. In this paper, we propose a diverse completion method that does not require a training set and can thus treat arbitrary images from any domain. Our internal diverse completion (IDC) approach draws inspiration from recent single-image generative models that are trained on multiple scales of a single image, adapting them to the extreme setting in which only a small portion of the image is available for training. We illustrate the strength of IDC on several datasets, using both user studies and quantitative comparisons.
Generative Adversarial Networks (GANs) have shown an outstanding ability to generate high-quality images with visual realism and similarity to real images. This paper presents a new architecture for thermal image enha...
详细信息
ISBN:
(纸本)9798350302493
Generative Adversarial Networks (GANs) have shown an outstanding ability to generate high-quality images with visual realism and similarity to real images. This paper presents a new architecture for thermal image enhancement. Precisely, the strengths of architecture-based vision transformers and generative adversarial networks are exploited. The thermal loss feature introduced in our approach is specifically used to produce high-quality images. Thermal image enhancement also relies on fine-tuning based on visible images, resulting in an overall improvement in image quality. A visual quality metric was used to evaluate the performance of the proposed architecture. Significant improvements were found over the original thermal images and other enhancement methods established on a subset of the KAIST dataset. The performance of the proposed enhancement architecture is also verified on the detection results by obtaining better performance with a considerable margin regarding different versions of the YOLO detector.
暂无评论