Computer Vision for Automated Inspection and Analysis

Challenge

• Human analysts unable to keep pace with the volume of imagery and video requiring review
• Manual inspection processes that are slow, inconsistent, and fatiguing
• Need for real-time or near-real-time visual analysis at edge locations

Approach

• Curate and annotate training datasets representative of mission-specific visual tasks
• Train and validate object detection, classification, and segmentation models
• Optimize models for deployment on target hardware including GPU servers and edge devices
• Implement human-in-the-loop workflows for model-assisted analysis with confidence scoring

Typical Outcomes

• Automated detection and classification reducing analyst review burden
• Consistent, repeatable analysis across large image and video datasets
• Real-time visual analytics capability deployable to edge and tactical environments

Procurement Paths

• NASA SEWP V and SEWP VI for GPU compute infrastructure and edge devices
• GSA MAS for computer vision development and integration services
• ITES-4H for supporting Army hardware requirements

Compare vehicles and ordering steps on How to buy.

Partner Technology Examples

• NVIDIA
• AMD
• Hewlett Packard Enterprise
• Dell Technologies