CV Brief · Thursday, 30 April 2026
CV Brief
Research & Papers
Makeup Attack Detection for Facial Recognition Systems
New framework tackles disguise makeup presentation attacks on facial recognition using attention-guided patch-based detection. Addresses a real security gap in deployed face systems where advanced cosmetics and prosthetics defeat standard defenses.
Read more →FruitProM-V2: Probabilistic Maturity Detection for Harvest
Replaces hard classification boundaries with probabilistic maturity estimation for fruits and vegetables, directly improving harvest timing accuracy. Practical system for agricultural CV pipelines needing robust ripeness detection.
Read more →RADIO-ViPE: Monocular Semantic SLAM Without Calibration
Open-vocabulary semantic SLAM operating on raw monocular RGB video without camera intrinsics or depth, enabling 3D grounding of natural language queries in dynamic scenes. Eliminates calibration burden for deployed SLAM systems.
Read more →Tools & Releases
Detecting Road Defects with Computer Vision: UAV and Dashcam Automation
Tutorial on building automated road defect detection using Roboflow, covering deployment strategies for both UAV imagery and dashcam video with automatic inspection report generation. Directly applicable to practitioners building production CV systems for infrastructure monitoring.
Read more →AI Evals Becoming New Compute Bottleneck in Production Pipelines
Analysis of how evaluation costs are now constraining AI model iteration cycles, shifting the bottleneck from training to validation. Critical for CV teams managing model validation budgets and deployment decisions.
Read more →Granite 4.1 LLMs: Architecture and Training Insights
Deep dive into Granite 4.1 model construction methodology, offering insights into large-scale model building practices. Relevant for CV practitioners working with multimodal models or considering foundation models for vision tasks.
Read more →Tutorials & Guides
Aerial Image Annotation: Dataset Quality for Drone-Based CV Models
Explores data annotation strategies specific to aerial/drone imagery for training robust models. Directly applicable to UAV monitoring, urban planning, and geospatial CV pipelines.
Read more →Getting Started in CV/ML
Camera Calibration Fundamentals: Distortion Correction for Accurate Systems
Guide to camera calibration and how optical distortion impacts CV accuracy. Essential reading for practitioners building stereo vision, 3D reconstruction, or any system requiring metric precision.
Read more →pHash deduplication for video crops: use Hamming distance ≤10 as your threshold. Too tight misses duplicates, too loose removes valid unique crops.
Quick Links
- Report of the 5th PVUW Challenge: Towards More Diverse Modalities in Pixel-Level
- Evaluating the Alignment Between GeoAI Explanations and Domain Knowledge in Sate
- Sample Selection Using Multi-Task Autoencoders in Federated Learning with Non-II
- MixerCA: An Efficient and Accurate Model for High-Performance Hyperspectral Imag