Latest AI and machine learning research in smoking & tobacco for healthcare professionals.
Multiple instance learning (MIL) is the standard approach for whole-slide image (WSI) classification...
Implementation of digital pathology leads to an increased number of whole slide images (WSIs). The l...
Visual RAG has offered an alternative to traditional RAG. It treats documents as images and uses vis...
Background: Previous machine learning models to intraoperatively predict the molecular status of gli...
Human behavioral and mental health outcomes arise from interactions among genetic, environmental, an...
Shadow detection is commonly formulated as a vision-driven dense prediction problem, where models re...
Text-guided inpainting has made image forgery increasingly realistic, challenging both SID and IFL. ...
Image hashing provides compact representations for efficient storage and retrieval but is inherently...
Vision Transformers (ViTs) achieve strong data-driven scaling by leveraging all-to-all self-attentio...
Generative models are increasingly used for protein design, but the lack of standardized evaluation ...
Open-vocabulary object detection (OVOD) aims to detect both seen and unseen categories, yet existing...
Quantum machine learning has emerged as a promising tool for pattern recognition, yet many audio-foc...
Mechanistic interpretability aims to reverse-engineer transformer computations by identifying causal...
Image-based Joint-Embedding Predictive Architecture (I-JEPA) offers a promising approach to visual s...
Mamba's recurrent state h_t is, by construction, a compressed summary of every token seen so far. Th...
Video snapshot compressive imaging (SCI) enables the reconstruction of dynamic scenes from a single ...
Text-to-image person re-identification (TI-ReID) relies on natural-language text description to retr...
Deploying tiny object perception on edge platforms is challenging because practical systems must sat...
Pathology foundation models (PFMs) have recently emerged as powerful pretrained encoders for computa...
CLIP-based person re-identification (ReID) methods aggregate spatial features into a single global \...
Physical adversarial patch attacks critically threaten pedestrian detection, causing surveillance an...