This paper studies model-inversion attacks, in which the access to a model is abused infer information about training data. Since its first introduction by~\cite{fredrikson2014privacy}, such attacks have raised serious concerns given that data usually contain privacy sensitive information. Thus far, successful only been demonstrated on simple models, as linear regression and logistic regression. Previous attempts invert neural networks, even ones with architectures, failed produce convincing...

Abstract Being capable of dealing with both electrical signals and light, artificial optoelectronic synapses are great importance for neuromorphic computing receiving a burgeoning amount interest in visual information processing. In this work, an synapse composed Al/TiN x O 2– /MoS 2 /ITO (H‐OSD) is proposed experimentally realized. The H‐OSD can enable basic voltage‐induced synaptic functions such as the long/short‐term plasticity moreover be electrically adjusted. response to light...

The unregulated use of LLMs can potentially lead to malicious consequences such as plagiarism, generating fake news, spamming, etc. Therefore, reliable detection AI-generated text be critical ensure the responsible LLMs. Recent works attempt tackle this problem either using certain model signatures present in generated outputs or by applying watermarking techniques that imprint specific patterns onto them. In paper, we show these detectors are not practical scenarios. particular, develop a...

Weakly supervised semantic segmentation (WSSS) with image-level labels is a challenging task. Mainstream approaches follow multi-stage framework and suffer from high training costs. In this paper, we explore the potential of Contrastive Language-Image Pre-training models (CLIP) to localize different categories only without further training. To efficiently generate high-quality masks CLIP, propose novel WSSS called CLIP-ES. Our improves all three stages special designs for CLIP: 1) We...

Abstract Intelligent perception means that with the assistance of artificial intelligence (AI)‐motivated brain, flexible sensors achieve ability memory, learning, judgment, and reasoning about external information like human brain. Due to superiority machine learning (ML) algorithms in data processing intelligent recognition, systems possess match or even surpass systems. However, built‐in these need work on dynamic irregular surfaces, inevitably affecting precision fidelity acquired data....

Abstract Developing electronic skins (e‐skins) that are comparable to or even beyond human tactile perception holds significant importance in advancing the process of intellectualization. In this context, a machine‐learning‐motivated micropyramid array bimodal (MAB) e‐skin based on capacitive sensing is reported, which enables spatial mapping applications (proximity and pressure) implemented via fringing iontronic effects, such as contactless measurement 3D objects contact recognition...

Abstract Neuromorphic hardware equipped with associative learning capabilities presents fascinating applications in the next generation of artificial intelligence. However, research into synaptic devices exhibiting complex behaviors is still nascent. Here, an optoelectronic memristor based on Ag/TiO 2 Nanowires: ZnO Quantum dots/FTO was proposed and constructed to emulate biological behaviors. Effective implementation behaviors, including long short-term plasticity,...

In self-supervised representation learning, a common idea behind most of the state-of-the-art approaches is to enforce robustness representations predefined augmentations. A potential issue this existence completely collapsed solutions (i.e., constant features), which are typically avoided implicitly by carefully chosen implementation details. work, we study relatively concise framework containing components from recent approaches. We verify complete collapse and discover another reachable...

As an emerging electronic device, a memristor facilitates the realization of neuromorphic computing systems. However, high-performance systems require not only robust analog memristors but also digital with complementary functions. Here, digital–analog integrated based on Al/ZnO NPs/CuO NWs/Cu is proposed and demonstrated. First, electrical properties CuO NWs device are investigated, which exhibit write-once-read-many-times (WORM) performance. Then, typical bipolar resistive switching...

Recent Transformer-based 3D object detectors learn point cloud features either from point- or voxel-based representations. However, the former requires time-consuming sampling while latter introduces quantization errors. In this paper, we present a novel Point-Voxel Transformer for single-stage detection (PVT-SSD) that takes advantage of these two Specifically, first use sparse convolutions efficient feature encoding. Then, propose (PVT) module obtains long-range contexts in cheap manner...

Neural network compression empowers the effective yet unwieldy deep convolutional neural networks (CNN) to be deployed in resource-constrained scenarios. Most state-of-the-art approaches prune model filter-level according "importance" of filters. Despite their success, we notice they suffer from at least two following problems: 1) The redundancy among filters is not considered because importance evaluated independently. 2) Cross-layer filter comparison unachievable since defined locally...

Training deep neural networks from scratch could be computationally expensive and requires a lot of training data. Recent work has explored different watermarking techniques to protect the pre-trained potential copyright infringements. However, these vulnerable watermark removal attacks. In this work, we propose REFIT, unified framework based on fine-tuning, which does not rely knowledge watermarks, is effective against wide range schemes. particular, conduct comprehensive study realistic...

Transformers have made great progress in dealing with computer vision tasks. However, existing transformers do not yet possess the ability of building interactions among features different scales, which is perceptually important to visual inputs. The reasons are two-fold: (1) Input embeddings each layer equal-scale, so no cross-scale feature can be extracted; (2) lower computational cost, some merge adjacent inside self-attention module, thus sacrificing small-scale (fine-grained) and also...

Existing neural rendering methods for creating human avatars typically either require dense input signals such as video or multi-view images, leverage a learned prior from large-scale specific 3D datasets that reconstruction can be performed with sparse-view inputs. Most of these fail to achieve realistic when only single image is available. To enable the data-efficient creation anima table humans, we propose ELICIT, novel method learning human-specific radiance fields image. Inspired by...

The Scene Graph Generation (SGG) task aims to detect all the objects and their pairwise visual relationships in a given image. Although SGG has achieved remarkable progress over last few years, almost existing models follow same training paradigm: they treat both object predicate classification as single-label problem, ground-truths are one-hot target labels. However, this prevalent paradigm overlooked two characteristics of current datasets: 1) For positive samples, some specific...