ResNet18
PulseAugur coverage of ResNet18 — every cluster mentioning ResNet18 across labs, papers, and developer communities, ranked by signal.
4 day(s) with sentiment data
-
New defense 'Combinatorial Lock' tackles multi-trigger AI backdoor attacks
Researchers have developed a novel defense mechanism called Combinatorial Lock to combat multi-trigger backdoor attacks in AI engines. This training-free approach treats the identification of multiple, heterogeneous att…
-
LiteMed-ViT: Lightweight Vision Transformer for Medical Imaging
Researchers have developed LiteMed-ViT, a lightweight Vision Transformer designed for medical image analysis on edge devices. This model uses a dual-token distillation approach to learn from a larger CNN teacher, signif…
-
AWS SageMaker SDK v3 streamlines edge computer vision model management
Amazon Web Services has updated its SageMaker SDK to version 3, introducing a new approach for managing computer vision models at the edge. This update replaces the older SageMaker Edge Manager with SageMaker Neo for op…
-
New TAN encoder processes histology images natively, reducing parameters and improving accuracy
Researchers have developed a new convolutional encoder called TAN (Tissue-Aware Network) designed to process histology images more effectively. Unlike previous methods that crop and pad superpixels, TAN processes each s…
-
LeanTTA framework enables quantized test-time adaptation on edge devices
Researchers have introduced LeanTTA, a novel framework designed for test-time adaptation of quantized machine learning models on edge devices. This backpropagation-free and stateless approach dynamically updates normali…
-
New deep learning patching method balances performance and training time
A new paper introduces Stride Independent Patching (SSP), a semi-automatic method for positioning image patches in deep learning. Unlike automatic stride-dependent methods, SSP relies on user input to define patch locat…
-
Drosophila connectome repurposed for AI tasks, showing promise for computer vision
Researchers have demonstrated that the neural connectome of Drosophila, a fruit fly, can be repurposed to perform human-designed cognitive tasks. By using the MaleCNS v1.0 connectome as a fixed network topology, models …
-
Paper explores hierarchy-aware deep learning for Tironian note recognition
A new paper explores the use of hierarchical deep learning models for recognizing Tironian notes, an ancient Latin shorthand system. Researchers compared standard classifiers like ResNet18 and Vision Transformers with h…
-
QSTAR framework enhances quantum transfer learning by selectively routing uncertain predictions
Researchers have developed QSTAR, a novel framework for quantum transfer learning that selectively routes uncertain predictions to a quantum branch. This approach aims to clarify the utility of quantum components in mac…
-
AI framework enhances Parkinson's disease screening using facial expression analysis
Researchers have developed FICAug, a novel framework designed to improve the screening of Parkinson's disease using facial expressions. This method addresses the challenge of small clinical datasets by employing feature…
-
New ProtoCAM framework improves few-shot breast lesion classification
Researchers have developed ProtoCAM, a novel interpretable few-shot learning framework designed for breast lesion classification in ultrasound imaging. This method integrates mask-guided feature encoding and prototypica…
-
New AI frameworks improve Alzheimer's diagnosis using MRI and clinical data
Researchers have developed new deep learning frameworks for diagnosing Alzheimer's disease using multimodal data. One study focuses on grounding image-based models with anatomical references and addressing label leakage…
-
New benchmark AAMBERS-UAV emphasizes acquisition-aware evaluation for drone weed segmentation
Researchers have developed a new benchmark called AAMBERS-UAV to evaluate multimodal backbone performance for weed segmentation in drone imagery. The study highlights the importance of acquisition-aware evaluation, whic…
-
New FLaG pooling method enhances AI model performance across domains
Researchers have introduced Frequency-Domain Latent-attention Gated Pooling (FLaG), a novel module designed to improve token aggregation by operating in the Fourier domain. This method re-expresses encoder outputs in th…
-
Deep learning system accurately identifies Bangladeshi mango varieties
Researchers have developed a deep learning-powered web system to identify Bangladeshi mango varieties, addressing the challenge of distinguishing similar cultivars. The system utilizes three fine-tuned CNN architectures…
-
Deep learning approach improves ovarian ultrasound classification accuracy
Researchers have developed a lesion-guided region-of-interest (ROI) deep learning approach for ovarian ultrasound classification, achieving high accuracy while reducing annotation effort. This method was evaluated on tw…
-
New framework offers causal interpretability for computer vision models
Researchers have developed a new causal interpretability framework for computer vision models, called Spatial Attention Noise Masking. This method dynamically masks input images before classification to provide causal e…
-
NPU Offloading Reduces Robot Training Energy at Cost of Time and Performance
Researchers have developed a method to reduce the energy consumption of training robot policies by offloading parts of the computation to a Neural Processing Unit (NPU). This approach involves freezing the visual encode…
-
Saliency-guided cutout shows mixed results for malware image classification
Researchers have investigated the effectiveness of saliency-guided cutout techniques for image-based malware classification. Their experiments, using the RawMal-TF dataset and ResNet18 architecture, compared four traini…
-
Classical SU(2) models outperform quantum circuits on vision tasks
A new research paper compares classical SU(2) models with variational quantum circuits (VQCs) on various vision benchmarks. The study found that quaternion-valued neural networks, a type of classical SU(2) model, perfor…