ImageNet
PulseAugur coverage of ImageNet — every cluster mentioning ImageNet across labs, papers, and developer communities, ranked by signal.
10 day(s) with sentiment data
-
New adaptive sparse coding framework improves visual signal robustness
Researchers have developed a new adaptive convolutional sparse coding (CSC) framework designed to create more robust visual signal representations. This method treats the sparsity coefficient as a learnable variable, op…
-
New pruning method considers channel response interactions for improved accuracy
Researchers have developed a novel approach to structured pruning in neural networks, moving beyond channel-ranking to consider the interactions between channel responses. This new method formulates pruning as selecting…
-
AI models improve Pap smear classification trustworthiness
Researchers have explored selective prediction and uncertainty-aware referral for Pap smear classification using deep learning models. By fine-tuning Swin-Tiny and TinyViT-5M transformers on the Herlev Pap smear dataset…
-
FPGA platform accelerates approximate multiplier evaluation for DNNs
Researchers have developed FAME, a new platform utilizing FPGAs to accelerate the evaluation of approximate multipliers for deep neural networks. This hardware-based approach significantly reduces the time needed to ass…
-
evMLP: New event-driven MLP architecture for efficient vision processing
Researchers have introduced evMLP, a novel all-MLP architecture designed for vision tasks, particularly video processing. This architecture leverages an event-driven mechanism that processes only changed regions between…
-
New benchmark GeoCrossBench targets cross-band generalization for remote sensing models
Researchers have introduced GeoCrossBench, an extension of the GeoBench benchmark designed to evaluate the cross-band generalization capabilities of remote sensing foundation models. This new benchmark includes protocol…
-
New method CoTS improves AI model calibration without sacrificing accuracy
Researchers have developed CoTS, a new post-hoc calibration method for test-time prompt tuning (TPT) that aims to improve accuracy without sacrificing calibration. CoTS applies temperature scaling to reduce the confiden…
-
Dual Randomized Smoothing enhances neural network robustness
Researchers have introduced Dual Randomized Smoothing (Dual RS), a novel framework designed to enhance the robustness of neural networks against adversarial perturbations. Unlike traditional Randomized Smoothing which u…
-
New deep learning benchmark for dental radiograph classification released
Researchers have developed a new deep learning benchmark for classifying periapical radiographs, addressing issues of patient-level data splits and cross-center validation. The benchmark, applied to the DentIRO dataset,…
-
Land Art Serves as AI-Powered Climate Indicator
Researchers have developed a novel method to use Robert Smithson's 1970 land artwork, Spiral Jetty, as a climate sensor by analyzing satellite imagery. By examining 1,744 image chips from 1984 to 2025, they extracted co…
-
New foundation model advances lunar remote sensing with multimodal capabilities
Researchers have developed a multimodal foundation model specifically for lunar remote sensing, leveraging a novel architecture and a large dataset called SoMBench. This model, named TerraMind, integrates data from vari…
-
New CertDW method offers certified dataset ownership verification
Researchers have developed CertDW, a novel method for verifying dataset ownership in deep neural networks. This approach uses conformal calibration to ensure reliable verification even when models are subjected to malic…
-
Logit Refiner enhances visual autoregressive models by restoring intra-scale dependencies
Researchers have developed a new method called the "Logit Refiner" to improve the quality of images generated by Visual Autoregressive Models (VAR). This technique addresses a limitation in VARs where parallel decoding …
-
New gradient inversion attack reveals significant privacy risks in federated learning
Researchers have developed a new method for gradient inversion attacks in federated learning, inspired by LT codes. This technique allows for exact recovery of training data and labels from a single round of FedSGD, sig…
-
New lightweight AI model AgroVisNet aids crop disease classification
Researchers have developed AgroVisNet, a lightweight convolutional neural network designed for plant disease classification on devices with limited connectivity and computational power. This model, along with the expert…
-
New framework uses foundation models for efficient communication
Researchers have introduced Foundation Model-Guided Semantic and Goal-Oriented Communication (FMSGOC), a new framework designed to improve generalization in communication systems, particularly for 6G applications. This …
-
New framework discovers natural transformation vulnerabilities in vision models
Researchers have developed a new framework called Adversarial Scenario Attack (ASA) to identify vulnerabilities in black-box vision models. ASA utilizes a multimodal language model and a text-guided generative editor to…
-
New JPPO attacks exploit vision-language models by optimizing pixels and prompts
Researchers have developed a new adversarial framework called Joint Pixel-Prompt Optimization (JPPO) that targets vision-language models (VLMs). Unlike previous methods that focused on image perturbations, JPPO jointly …
-
ImageNet augmentation rivals massive datasets for text-to-image generation
Researchers have demonstrated that text-to-image generation models can achieve high performance using only the ImageNet dataset, augmented with text and image enhancements. This approach challenges the prevailing 'bigge…
-
New attention framework enhances GI endoscopy image classification
Researchers have developed MultiAttenGastro, a novel attention framework designed to improve the classification of gastrointestinal endoscopy images. The framework employs parallel 1-D, 2-D, and 3-D attention heads to c…