Advanced Neural Network Applications
New papers in advanced neural network applications, the code and data they release, and the open problems their authors admit. Rebuilt every day from OpenAlex and arXiv.
New papers
Supernetwork Architecture Selection for Deployable Language Models: RAG Assessment of Overhead-Line Insulators on a 2 GB Device for Grid Operator's Maintenance Crews
Transmission system operators inspecting high-voltage insulator strings using inference using drone capture inference in remote locations need a local LLM assistant that is useful only if its answers are hallucination-free, drawn from operator-held documents and never exceeding what a photograph…
ConvRoute-ViT: Convolutional Token Routing for Efficient Vision Transformers
In this paper, we propose ConvRoute-ViT, a convolutional token routing mechanism for efficient vision transformers that performs token reduction through convolutional routing.
TRIM: Temporal Run-Time Monitoring of 3D Object Detection via Activation Statistics
We present TRIM, a lightweight temporal monitor that detects missed-object failures without modifying or retraining the underlying detector.
SaiVLA-JEPA: Evaluating Future Latent Prediction in JEPA-Conditioned VLA Control
Joint-embedding predictive architectures (JEPAs) learn to predict representations rather than pixels, offering a compact way to model how visual scenes evolve.
Lookahead Guidance Intervention Optimization for Diffusion Models: Predicting Downstream Guidance Effects to Select Adaptive Conditioning Strategies
This paper proposes Lookahead Guidance Intervention Optimization (LGIO), a conceptual framework for selecting conditioning interventions according to their predicted downstream effects on final generation quality.
A Survey of AI-Based Railway Track Object Detection and Alert Systems
Railway safety is affected by the intrusion of people, animals, vehicles, debris, and other foreign objects into railway track areas.
Seeing Through Draw Calls: Learning Object Semantics from a Game's Render Commands
AI systems that need to understand a game or 3D application today look at its pixels.
Household Gas Facility Detection Using RT-DETR with Morphology-Aware Feature Enhancement and Short-Axis-Aware Regression
This study presents an enhanced RT-DETR-R18 detector combining morphology-aware feature enhancement, local–global interaction, and geometry-sensitive regression.
Lookahead Model Invocation Intervention Optimization: Predicting the Downstream Value of Auxiliary Model Calls for Budget-Aware Diffusion Generation
This paper proposes Lookahead Model Invocation Intervention Optimization (LMIO), a conceptual framework for deciding whether, when, and how to invoke auxiliary models according to their predicted downstream value.
Spatial contextual learning for hyperspectral imagery classification using transformer encoder framework based ensemble model
In this work, we propose a four–stage hybrid framework for hyperspectral image classification.
Lookahead Objective Reweighting Intervention Optimization for Diffusion Models: Predicting Downstream Generation Outcomes to Select Adaptive Multi-Objective Priorities
This paper proposes Lookahead Objective Reweighting Intervention Optimization (LORIO), a conceptual framework that selects objective-weight configurations according to their predicted downstream generation outcomes.
Lookahead Multiscale Fusion Intervention Optimization: Predicting Downstream Generation Quality to Select Adaptive Cross-Scale Feature Fusion Decisions
This paper proposes Lookahead Multiscale Fusion Intervention Optimization (LMFIO), a conceptual framework for selecting cross-scale feature-fusion interventions according to their predicted effects on future generation quality.
Lookahead Conditioning-Encoder Layer Intervention Optimization (LCELIO): Predicting Downstream Generation Outcomes to Select Adaptive Layer-Level Conditioning Interventions
This paper proposes Lookahead Conditioning-Encoder Layer Intervention Optimization (LCELIO), a conceptual framework for selecting conditioning-encoder layer interventions according to their predicted effects on future generation outcomes.
DFSMamba: Detail-Guided and Direction-Adaptive State-Space Modeling for Structure-Preserving Visible-to-Infrared Image Translation
We propose DFSMamba to preserve geometric structure throughout visible-to-infrared translation by connecting frequency-guided feature formation, input-dependent directional aggregation, and infrared-domain reconstruction supervision.
Leakage-Aware Road Damage Detection Based on GAM-YOLOv8
To address these issues, this paper proposes a leakage-aware evaluation protocol and develops a lightweight road damage detection method that combines class rebalancing with global attention enhancement.
The Architecture of Permanence: Deterministic Edge Governance Across Heterogeneous NVIDIA Robotics on Commodity Silicon
Executive Overview & Problem Context Current Vision-Language-Action (VLA) foundation models rely heavily on brute-force scale, empirical loss minimization, and post-hoc alignment (such as RLHF).
Lookahead Constraint Activation Intervention Optimization for Diffusion Models: Predicting Downstream Generation Quality to Adaptively Schedule Structural Constraints
This paper proposes Lookahead Constraint Activation Intervention Optimization (LCAIO), a conceptual framework that selects structural-constraint interventions according to their predicted downstream effects on final generation quality.
Lookahead Latent Quantization Intervention Optimization (LLQIO): Predicting Downstream Generation Quality to Select Adaptive Precision in Diffusion Models
This paper proposes Lookahead Latent Quantization Intervention Optimization (LLQIO), a conceptual framework for selecting quantization interventions according to their predicted downstream effects on generation quality and computational efficiency.
Online Detection of Molten Pool State in Flame Brazing Based on YOLOv12 and Kalman Filter
To address the highly variable molten pool morphology, intense flame and smoke interference, and strict online detection requirements in flame brazing, this paper proposes a welding state detection method based on YOLOv12 and a Kalman filter.
Smart Vehicle Identification and Tracking Using YOLO-Based License Plate Recognition
Manual verification of vehicle registration plates can be time consuming and error-prone.
Released code and data
Asymmetric encoder-decoder U-net with hybrid attention and residual dilated convolution for brain tumor segmentation
Results The proposed model was evaluated on the publicly available BRISC 2025 and Figshare datasets, achieving Dice coefficients of 87.39% and 82.84%, respectively Discussion These findings support the complementary roles of the HCSA and RMSDC mechanisms in enhancing feature representation and multi-scale contextual modeling.
GeoFormer: Non-Equilibrium Metric Manifolds and Quantized Thermodynamic Attractor Convergence for Multi-Hop Transitive Reasoning
Open Source Code & Repositories GitHub Repository: https://github.com/bhosaleone/geoformer Hugging Face Weights: https://huggingface.co/SetMD/geoformer Author Research Dispatches: https://praman.blog Licensing & Commercial Rights GeoFormer is released under a Dual-Licensing Model: Open-Source Research: Software and algorithms are licensed under the GNU Affero General Public License v3.0 (AGPLv3).
Inference Fabric: A throughput-first GPU inference pipeline on consumer NVIDIA hardware: broker, saturated runtime, completion, vector store
Repository document: https://github.com/coreyjames-ai/inference-fabric, mirrored at https://gitlab.com/coreyjames-ai/inference-fabric.
IV-PAFNet: Infrared-visible cross-modal alignment and fusion network for nighttime UAV object detection
The code is available at https://github.com/iv-pafnet/IV-PAFNet .
A cascaded parallel deep learning framework for precise potato canopy density mapping via edge-aware feature fusion
The training program can be obtained from https://github.com/HT-JL/Potato-Research.git .
ICTNet: Independent CNN–Transformer Dual-Branch Network for Poppy Object Detection
Moreover, we create a multi-scene poppy detection dataset named MS-Poppy, which contains 2443 images covering six UAV-view scenes.
Robust Edge-AI for Explainable Maize Leaf Disease Diagnosis using CNN-Transformer Fusion
Model checkpoints, code, and Android binaries are open-sourced at GitHub - muslimraza9989-cmyk/Asan-Kissan-EdgeAI
A lightweight C Code Generator for Heterogeneous TinyML Systems
The code is open-sourced in Gitlab.
Nonlinear Direct Feedback Alignment for Scalable Backpropagation-Free Training
The code is available at https://github.com/YanFan321/NDFA.
Frugal Handwritten Text Recognition for Sub-Saharan Civil-Status Registers: A Smartphone-to-Records Pipeline with CPU-Only Inference
We introduce a corpus of 1,568 smartphone images of handwritten birth, marriage, and death records from a municipal registry in Chad, with line-level transcriptions for 185 records.
Open problems authors admit
The Architecture of Permanence: Deterministic Edge Governance Across Heterogeneous NVIDIA Robotics on Commodity Silicon
Future Work: Plans focus on expanding topological invariance to high-degree-of-freedom humanoid systems (e.g., NVIDIA Project GROOT), extending the invariant anchor manifold across 20+ simultaneous joint limits, whole-body center-of-mass (CoM) stability polygons, and zero-moment point (ZMP) dynamic constraints.
Deep Learning Approaches for Skin Cancer Detection: A Systematic Literature Review on Architectures, Multimodal Fusion Strategies, and Open Research Challenges
The review concludes that while reported accuracy in this field continues to rise, trustworthy evaluation methodology has not kept pace, and outlines concrete directions for future work.
RSDMF-DETR: A Remote Sensing Direction-aware Multi-scale Fusion Detection Transformer
Its deployment and performance on actual UAV or edge-computing platforms have not yet been experimentally validated and remain topics for future work.
Deep learning for optical music recognition: A review
Additionally, it explores emerging trends, identifies key challenges, and proposes future directions to advance OMR research.
ACCM-Net: an adaptive CLAHE-CBAM-MobileNet architecture for wheat leaf disease classification
Future work will extend the model to other disease classes and explore transformer-based enhancements to further contextual understanding.
Multi-Weather Classification Using Machine Learning: A Comparative Study
the dataset, including class distribution and feature characteristics. • Openly available source code and notebooks to ensure full reproducibility of the results.This work serves as a solid foundation for future research and practical applications in environmental and weather-related computer vision tasks.
SLICE-CiM: Selective Layer-wise Inference Control Engine with Confidence-Based Early Exit for SRAM based Digital Compute-in-Memory
The framework targets CNN workloads exploiting sparsity, with Transformer extension as future work.
Progressively Refining SAM for Accurate Phytoplankton Segmentation
In addition, we construct a phytoplankton segmentation dataset by manually annotating instance-level masks for an existing dataset, providing a valuable resource for future research.
uTHCD-C: A Corruption-Robustness Benchmark for Tamil Handwritten Character Recognition
By providing standardized corruption configurations, evaluation protocols, and comparative results, uTHCD-C establishes a reproducible basis for future research on robust and computationally efficient Tamil HCR.
Attention Educosmos: A Smart Classroom Focus and Smartphone Monitoring System
We report the parameters and the measurements behind them from a single-classroom recording, and we state plainly that a formal precision/recall evaluation on a larger dataset remains future work.