ResNet-50
PulseAugur coverage of ResNet-50 — every cluster mentioning ResNet-50 across labs, papers, and developer communities, ranked by signal.
8 day(s) with sentiment data
-
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…
-
New watermarking scheme TwinMark protects AI models from distillation attacks
Researchers have developed TwinMark, a novel watermarking technique designed to protect AI models against distillation attacks. This method uses two complementary linear functionals, one based on feature covariance and …
-
Deep learning models enhance wildfire spread prediction accuracy and auditability
Two new research papers explore the application of deep learning models for predicting wildfire spread. The first paper, focusing on the Rectoret region in Spain, compares four architectures including U-Net, ResNet-50, …
-
2D backbone choice significantly impacts AI's 3D spatial understanding
A new research paper explores the impact of different 2D image backbones on indoor semantic occupancy prediction. The study found that the choice of backbone significantly influences the accuracy of 3D predictions, more…
-
New benchmark dataset launched for lunar science machine learning
Researchers have introduced SoMBench, a new benchmark dataset designed to advance machine learning applications in lunar science. This dataset unifies data from over ten instruments across four lunar missions, including…
-
New variational template matching framework improves anomaly detection in structured images
Researchers have developed a new variational template matching framework for anomaly detection in structured images, particularly effective in small-data scenarios where deep learning is impractical. This method represe…
-
AI predicts optimal build orientation for 3D-printed dental parts
Researchers have developed a machine learning approach to predict the optimal build orientation for dental parts manufactured using selective laser melting (SLM). By training models on approximately 2400 patient-specifi…
-
New machine unlearning method achieves 82x speedup
Researchers have developed a novel machine unlearning framework that significantly speeds up the process of removing specific data points from trained models. This method identifies correlated data points and uses a clo…
-
Phase transition frequency predicts ResNet accuracy in training
Researchers have identified a new metric, "phase transition frequency," that can predict the test accuracy of ResNet models during training. This metric, which counts discrete class-separability jumps, showed a strong n…
-
New method tackles rotation-induced drift in AI model explanations
Researchers have identified a significant issue with post-hoc saliency maps, such as Grad-CAM, used to audit AI model decisions. These maps exhibit a 'drift' when input images are rotated, even if the model's prediction…
-
Dermatology AI generalization gap linked more to disease shift than skin tone
A new study published on arXiv investigates the generalization gap in dermatology AI models, specifically examining whether poor performance is due to underrepresentation of skin tones or shifts in disease distribution.…
-
New benchmark evaluates vehicle attribute classification in surveillance
Researchers have introduced the Unconstrained Vehicle Identification Benchmark (UVIB) to evaluate vehicle attribute classification in diverse surveillance scenarios. This benchmark, comprising 84,835 images from seven B…
-
New VQA model automates NDE image analysis using ResNet-50 and GPT-2
Researchers have developed a Visual Question Answering (VQA) model tailored for nondestructive evaluation (NDE) image analysis. This system integrates a ResNet-50 model for image feature extraction and GPT-2 for languag…
-
New deep learning model enhances peatland fire detection accuracy
Researchers have developed a new deep learning framework, WHT-ResNet-50, designed for more accurate and efficient detection of peatland fires. This model utilizes a Walsh-Hadamard Transform to enhance feature representa…
-
Human crop decisions outperform AI scaling for book digitization
Researchers have developed a novel method for automating the digitization of rare books by leveraging a decade's worth of manual crop decisions made by human operators. This approach, which involved recovering 575,729 c…
-
New audit method reveals true interactions in language models
Researchers have developed a new method called a site-asymmetry audit to better understand interactions within neural networks. This audit helps distinguish genuine interaction effects from those caused by the location …
-
New Nepali-English benchmark for misinformation detection released
Researchers have developed NepOOC-M, the first publicly available benchmark for detecting out-of-context (OOC) misinformation in Nepali and English. The dataset includes 1,090 image-caption pairs annotated across five t…
-
New AnchorScore method predicts MLLM annotation difficulty using CLIP
Researchers have developed AnchorScore, a novel method utilizing CLIP to predict the difficulty multimodal large language models (MLLMs) face in annotating specific classes. This approach offers a low-cost diagnostic to…
-
New research explores set decoder performance in computer vision
Researchers have developed a new method for analyzing set decoders in computer vision, focusing on the tension between improving individual predictions and maintaining the overall utility of the prediction set. Their st…
-
AI model-brain comparisons sensitive to image resolution, study finds
A new study published on arXiv investigates how the resolution at which convolutional neural networks are evaluated can significantly impact comparisons between different learning rules, particularly in the context of m…