Thermo-VL augments a frozen Molmo-7B VLM with a trainable thermal encoder and prompt-conditioned dual-attention fusion to improve cross-spectrum visual reasoning.
hub Canonical reference
An Introduction to Convolutional Neural Networks
Canonical reference. 83% of citing Pith papers cite this work as background.
23
Pith papers citing it
Background
83% of classified citations
abstract
The field of machine learning has taken a dramatic twist in recent times, with the rise of the Artificial Neural Network (ANN). These biologically inspired computational models are able to far exceed the performance of previous forms of artificial intelligence in common machine learning tasks. One of the most impressive forms of ANN architecture is that of the Convolutional Neural Network (CNN). CNNs are primarily used to solve difficult image-driven pattern recognition tasks and with their precise yet simple architecture, offers a simplified method of getting started with ANNs. This document provides a brief introduction to CNNs, discussing recently published papers and newly formed techniques in developing these brilliantly fantastic image recognition models. This introduction assumes you are familiar with the fundamentals of ANNs and machine learning.
hub tools
citation-role summary
background 6
citation-polarity summary
roles
background 6representative citing papers
OVS-DINO structurally aligns DINO with SAM to revitalize attenuated boundary features, achieving SOTA gains of 2.1% average and 6.3% on Cityscapes in weakly-supervised open-vocabulary segmentation.
SpectraLLM is an LLM fine-tuned to predict small-molecule structures from single or multiple spectra, reporting state-of-the-art results on four public benchmarks with gains from multi-modal input.
QMC-Net maps per-band statistics to customized quantum circuit hyperparameters and achieves 93.80% and 99.34% accuracy on EuroSAT and SAT-6, outperforming classical and monolithic quantum baselines.
AP-MAE reconstructs masked attention patterns in LLMs with high accuracy, generalizes across models, predicts generation correctness at 55-70%, and enables 13.6% accuracy gains via targeted interventions.
A recursive cubing framework identifies stable hyperparameter regions for MC dropout uncertainty quantification in spatial deep learning and produces competitive or superior predictive intervals versus a statistical baseline on simulations and land-surface temperature data.
AlayaLaser uses a SIMD-optimized on-disk graph layout plus caching and search strategies to outperform prior on-disk ANNS systems and match or exceed in-memory performance on large high-dimensional datasets.
Short-time averages within experiments plus temporal-preserving models like CNNs cut multiphase mass flow metering errors to 4.3% MAPE on air-water-oil data, outperforming single-averaged baselines.
Genome-Factory is an open-source Python library that integrates data pipelines, model tuning, inference, benchmarks, and biological interpretation for genomic foundation models.
AdaProb performs machine unlearning by substituting final-layer output probabilities with optimized uniform pseudo-probabilities and updating model weights.
Using a deep CNN and Fourier frequency analysis on calorimeter data, the KOTO experiment suppressed neutron background by a factor of 5.6×10^5 while maintaining 70% efficiency for the signal decay.
The work introduces WaLeF/FIDLAr for flood forecasting, CoDiCast for probabilistic weather, and Hypercube-RAG for explainable environmental QA, claiming superior accuracy, efficiency, and interpretability over baselines.
LRCN and Transformer models using GelSight tactile images improve compliance prediction accuracy over baselines and show that objects harder than the sensor are harder to estimate.
HeartBERT applies self-supervised pretraining on a RoBERTa architecture to ECG signals, producing embeddings that enable strong performance on sleep staging and heartbeat classification with smaller labeled datasets and fewer parameters than baselines.
TemPose-TF-ASF fuses bidirectional stroke context in two stages to raise accuracy and Macro-F1 for badminton stroke classification over baselines.
SAGE-GAN integrates a self-attention U-Net into a CycleGAN framework to generate realistic synthetic electron microscopy image-mask pairs that augment training data for nanoparticle segmentation without human labeling.
Pre-trained ViT representations combined with active learning and targeted design choices for annotations and selection improve object class retrieval in multi-object scenes.
EGI integrates four existing AI components for real-time multimodal emotion monitoring and feedback in simulated agile meetings, reporting 10% WER and improved self-awareness for Scrum Masters.
HuBERT reaches 86% accuracy and 0.93 AUC detecting COVID-19 from 893 voice samples in the Cambridge COVID-19 Sound database.
The paper overviews attention-based learning methods for spectrum cartography in LEO satellite networks to enable adaptive fusion of heterogeneous measurements for inference and resource allocation.
A survey providing an overview of publicly available image-based datasets for ML/DL-based disaster management pipelines covering pre-disaster, during, and post-disaster phases.
A holistic survey of affective computing for intelligent agents covering emotion understanding via multimodal data, affective cognition, emotional expression synthesis, key challenges, and future directions emphasizing generative technologies.
A 75 ms Gaussian window for segmenting phonocardiography signals yields the highest biLSTM classification accuracy among tested shapes and lengths, outperforming rectangular windows and a baseline method.
citing papers explorer
-
Thermo-VL: Extending Vision-Language Models to Thermal Infrared Perception
Thermo-VL augments a frozen Molmo-7B VLM with a trainable thermal encoder and prompt-conditioned dual-attention fusion to improve cross-spectrum visual reasoning.
-
OVS-DINO: Open-Vocabulary Segmentation via Structure-Aligned SAM-DINO with Language Guidance
OVS-DINO structurally aligns DINO with SAM to revitalize attenuated boundary features, achieving SOTA gains of 2.1% average and 6.3% on Cityscapes in weakly-supervised open-vocabulary segmentation.
-
SpectraLLM: Uncovering the Ability of LLMs for Molecular Structure Elucidation from Multi-Spectral Data
SpectraLLM is an LLM fine-tuned to predict small-molecule structures from single or multiple spectra, reporting state-of-the-art results on four public benchmarks with gains from multi-modal input.
-
QMC-Net: Data-Aware Quantum Representations for Remote Sensing Image Classification
QMC-Net maps per-band statistics to customized quantum circuit hyperparameters and achieves 93.80% and 99.34% accuracy on EuroSAT and SAT-6, outperforming classical and monolithic quantum baselines.
-
Automated Attention Pattern Discovery at Scale in Large Language Models
AP-MAE reconstructs masked attention patterns in LLMs with high accuracy, generalizes across models, predicts generation correctness at 55-70%, and enables 13.6% accuracy gains via targeted interventions.
-
A Cubing Strategy for Identifying Stable Hyperparameter Regions for Uncertainty Quantification in Spatial Deep Learning
A recursive cubing framework identifies stable hyperparameter regions for MC dropout uncertainty quantification in spatial deep learning and produces competitive or superior predictive intervals versus a statistical baseline on simulations and land-surface temperature data.
-
AlayaLaser: Efficient Index Layout and Search Strategy for Large-scale High-dimensional Vector Similarity Search
AlayaLaser uses a SIMD-optimized on-disk graph layout plus caching and search strategies to outperform prior on-disk ANNS systems and match or exceed in-memory performance on large high-dimensional datasets.
-
Temporal Data and Short-Time Averages Improve Multiphase Mass Flow Metering
Short-time averages within experiments plus temporal-preserving models like CNNs cut multiphase mass flow metering errors to 4.3% MAPE on air-water-oil data, outperforming single-averaged baselines.
-
Genome-Factory: A Library for Tuning, Deploying, and Interpreting Genomic Foundation Models
Genome-Factory is an open-source Python library that integrates data pipelines, model tuning, inference, benchmarks, and biological interpretation for genomic foundation models.
-
AdaProb: Efficient Machine Unlearning via Adaptive Probability
AdaProb performs machine unlearning by substituting final-layer output probabilities with optimized uniform pseudo-probabilities and updating model weights.
-
Suppression of Neutron Background using Deep Neural Network and Fourier Frequency Analysis at the KOTO Experiment
Using a deep CNN and Fourier frequency analysis on calorimeter data, the KOTO experiment suppressed neutron background by a factor of 5.6×10^5 while maintaining 70% efficiency for the signal decay.
-
Accurate, Efficient, and Explainable Deep Learning Approaches for Environmental Science Problems
The work introduces WaLeF/FIDLAr for flood forecasting, CoDiCast for probabilistic weather, and Hypercube-RAG for explainable environmental QA, claiming superior accuracy, efficiency, and interpretability over baselines.
-
Advances in Compliance Detection: Novel Models Using Vision-Based Tactile Sensors
LRCN and Transformer models using GelSight tactile images improve compliance prediction accuracy over baselines and show that objects harder than the sensor are harder to estimate.
-
HeartBERT: A Self-Supervised ECG Embedding Model for Efficient and Effective Medical Signal Analysis
HeartBERT applies self-supervised pretraining on a RoBERTa architecture to ECG signals, producing embeddings that enable strong performance on sleep staging and heartbeat classification with smaller labeled datasets and fewer parameters than baselines.
-
TemPose-TF-ASF: Two-Stage Bidirectional Stroke Context Fusion for Badminton Stroke Classification
TemPose-TF-ASF fuses bidirectional stroke context in two stages to raise accuracy and Macro-F1 for badminton stroke classification over baselines.
-
SAGE-GAN: Towards Realistic and Robust Segmentation of Spatially Ordered Nanoparticles via Attention-Guided GANs
SAGE-GAN integrates a self-attention U-Net into a CycleGAN framework to generate realistic synthetic electron microscopy image-mask pairs that augment training data for nanoparticle segmentation without human labeling.
-
Revisiting Human-in-the-Loop Object Retrieval with Pre-Trained Vision Transformers
Pre-trained ViT representations combined with active learning and targeted design choices for annotations and selection improve object class retrieval in multi-object scenes.
-
EGI: A Multimodal Emotional AI Framework for Enhancing Scrum Master Real-time Self-Awareness
EGI integrates four existing AI components for real-time multimodal emotion monitoring and feedback in simulated agile meetings, reporting 10% WER and improved self-awareness for Scrum Masters.
-
Developing a Multi-variate Prediction Model For COVID-19 From Crowd-sourced Respiratory Voice Data
HuBERT reaches 86% accuracy and 0.93 AUC detecting COVID-19 from 893 voice samples in the Cambridge COVID-19 Sound database.
-
Learning-Based Spectrum Cartography in Low Earth Orbit Satellite Networks: An Overview
The paper overviews attention-based learning methods for spectrum cartography in LEO satellite networks to enable adaptive fusion of heterogeneous measurements for inference and resource allocation.
-
Survey on Disaster Management Datasets for Remote Sensing Based Emergency Applications
A survey providing an overview of publicly available image-based datasets for ML/DL-based disaster management pipelines covering pre-disaster, during, and post-disaster phases.
-
Intelligent Agents with Emotional Intelligence: Current Trends, Challenges, and Future Prospects
A holistic survey of affective computing for intelligent agents covering emotion understanding via multimodal data, affective cognition, emotional expression synthesis, key challenges, and future directions emphasizing generative technologies.
-
Comparison of window shapes and lengths in short-time feature extraction for classification of heart sound signals
A 75 ms Gaussian window for segmenting phonocardiography signals yields the highest biLSTM classification accuracy among tested shapes and lengths, outperforming rectangular windows and a baseline method.