The Problem with Manual Digitization
Every oil and gas company has the same problem: filing cabinets (or file servers) filled with raster well log images. These TIFF and PDF scans contain critical subsurface data - SP, GR, resistivity, sonic, density curves - but the data is trapped in pixels. You cannot query it, plot it, or feed it into modern petrophysical software without first converting it to digital LAS format.
Manual digitization - where a geologist traces each curve by hand using desktop software - takes 4 to 6 hours per log. For a company with 10,000 well logs, that is 40,000 to 60,000 hours of work. At any reasonable billing rate, the cost is prohibitive.
How Nspect Solves This
Nspect uses computer vision and specialized neural networks trained on hundreds of thousands of well log images to automatically:
- Detect track boundaries - identifies the grid structure, depth scale, and individual log tracks
- Trace curves - extracts each curve's path from the raster image, distinguishing between SP (typically on the left), resistivity curves (middle tracks), and other measurements
- Assign values - maps pixel positions back to engineering units using the detected scale headers
- Export to LAS - generates industry-standard LAS 2.0 files ready for import into any petrophysical package
The entire process takes under 30 seconds per log image - a 480x speedup over manual methods.
Quality Control Built In
Speed without accuracy is worthless. That is why Nspect includes a full interactive QC environment:
- Side-by-side overlay - the digitized curves are rendered on top of the original raster image so you can visually verify accuracy
- Joiner and Razor tools - fix curve breaks and trim artifacts with keyboard shortcuts
- Scale editor - adjust depth scales and curve scales interactively
- Multi-project management - organize thousands of images across multiple projects with status tracking
Each image moves through a status pipeline: Pending, In Progress, QC Review, Complete. Batch operations let you process entire wells at once, and batch LAS export generates one file per well with all curves merged.
Real-World Performance
In production deployments, Nspect consistently achieves:
- 95%+ accuracy on clean raster scans with standard grid layouts
- 85-90% accuracy on degraded scans with smudges, folds, or non-standard grids
- 7 curve types extracted simultaneously: SP, GR, SFL, ILD, ILM, DT, CILD
The remaining edge cases are handled by the QC workflow. An operator who previously spent weeks on a batch of 500 logs now completes the same batch in a single day - including full QC review.
Deployment
Like all NthDS products, Nspect runs on-premise with local GPU inference. Your well log data stays on your servers. The platform is accessed through a web browser, requires no desktop installation, and supports concurrent multi-user workflows.
Want to see Nspect process your logs? Request a demo.