Top 100 Blind Spots — Packaged Datasets
We converted the Top 100 AI blind spots into discrete dataset packages. Each package contains curated sensor streams, metadata, labeling guidance and suggested evaluation tasks so researchers and developers can close critical gaps.
You can explore a sample subset below. The full list is available on request or via our enterprise catalog.
Human-Centric: Emotion Recognition Pack
ECG, facial EMG proxies, audio prosody and GSR with emotion labels.
Environmental: Hyperlocal Air Quality
Dense PM2.5, PM10 and NO₂ sensors across micro-neighborhoods.
Mechanical: Pre-failure Vibration Signatures
High-frequency accelerometer captures labeled by failure type.
Mobility: Urban LiDAR & GNSS
Synchronized LiDAR point clouds with ground truth trajectories.
Energy: Grid Frequency Anomalies
High-resolution frequency and phase data for grid stability studies.
Agriculture: Soil Moisture Time Series
Multi-depth moisture sensors with irrigation and yield annotations.
Security: Directional Audio Events
Multi-microphone arrays with labeled gunshot, glass break and siren events.
Space: Radiation Flux Time Series
Satellite radiation sensors with solar event labeling.
Social: Crowd Density & Flow
Anonymized presence, flow and CO₂ proxies for public space analytics.
Visual & Sensory: Thermal Imagery
Infrared time-series aligned with annotation masks for thermal perception tasks.
Economic: Retail Shelf Sensors
Weight/pressure sensors, door-open events and environmental proxies for store analytics.
Explore Full Top 100 Catalog (Enterprise)
Full catalog access requires registration and approval — request access to get complete metadata, sample downloads and licensing options.