Skip to main content

Clarity updates since 0.100.0

Scope: Changes after 0.100.0 through production version 0.117.19, grouped by workflow. This is a cumulative update, not a list of features introduced only in 0.117.19. Access to Alpha features and Clarity AI depends on your organization's configuration.

Import and organize imagery​

  • Review uploaded files in a file inspection and metadata editing workspace. Import failures are surfaced after upload, with spectral-axis unit controls and Clarity-assisted intake and recovery.
  • Retain unrecognized image attachments and pair validated IGM files for sensor-grid imagery. Import and analysis validity checks exclude unusable spectral bands rather than treating them as measured data.
  • Organize canonical images through project and folder references without duplicating their underlying imagery. Catalog locations are optional, and folder rows can open their folders directly.
  • Use project-aware navigation across accounts you can access. Signed-out navigation and tab-local authentication recovery have also been corrected.

Explore, compare, and label​

  • Browse sources with search, sorting, processing-level filters, and a date timeline. Canvas filters can restrict the imagery and labels shown alongside the source list; source visibility controls preserve individual layer choices.
  • Inspect source links and thumbnails with clearer loading feedback. Native threshold previews follow the current view, and raster exports preserve validity masks and sparse pixels during reprojection.
  • Review image alignment on the map and publish accepted translations as independent image versions. Positioning reviews and ground checkpoint assessments retain their own review context.
  • Use project label collections, bulk selection and archiving, and unified class, copy, and delete actions. Labeling fixes preserve the current view and reduce conflicts between drawing and map dragging.
  • Material identification returns a ranked shortlist of up to five matches, including results for vector labels. Spectral libraries preserve imported chapters and support named groups, chapter management, and refreshed search after saving signatures.
  • Analysis additions include SAM2 discovery, N-FINDR prerequisites, tiled Dense CRF for large images, mask-to-label conversion, and layer round trips.

Prepare datasets and run models​

  • Merge source classes into managed output classes within a draft dataset version. Original annotations remain unchanged, and finalized versions retain their mapping. See Dataset labels.
  • Dataset preparation improves invalid-band handling, frozen bucket behavior, validation performance, and failure reporting. Wavelength details are easier to inspect in the dataset dialog.
  • Review per-image wavelength compatibility before inference. Unsupported inputs are rejected by default; explicit, bounded gap interpolation and edge extension are available when you choose to estimate missing wavelengths. Partial batches require confirmation and leave incompatible images skipped. See Inference results.
  • New model clones align with the selected dataset spectrum. Training resume checks, regression transfer inputs, model downloads, and model-test artifact staging have been corrected. Model inference results also gain a gallery.

Clarity AI, studies, and stories​

  • Study files use a folder tree and can open automatically from the conversation. Studies gain project ownership, deep links, pins, and file editing, with clearer navigation and more reliable refresh after rewrites.
  • Clarity AI can launch and track trained-model inference, use label-table context, and help recover imports. Follow-up conversations, live media, background completion delivery, and task failure reporting are more reliable.
  • Token usage reporting includes per-part counts and corrected pricing and context-window reporting. Delegated runs retain their own child sessions and configuration history.
  • Authored Story views support linked comparison layouts, independent scene visibility and camera bounds, and geographic scales. Mobile navigation, panel resizing, research sidebars, and workflow links have been refined.
  • Opt-in browser push notifications can report job completion when configured.

Learn more​