PanorAi 3.8.0

PanorAi 3.8.0 introduces a composable spherical computer-vision foundation while preserving the 3.x compatibility surface.

Highlights

• Adds the Stable typed Pipeline and spherical observation workflow for composing dense rasters, points, regions, and whole-view evidence.
• Adds Stable NumPy spherical operations and generic CPU Torch layers for direct ERP processing, including spherical convolution, pooling, resizing, gradients, and multiscale composition.
• Expands EquirectangularImage as the multimodal ERP container and adds explicit multi-view projection, sampling, processing, and fusion paths.
• Adds Stable spherical SLIC with seam-, support-, and solid-angle-aware behavior.
• Adds role-based component documentation and a machine-readable public API catalog so Stable, Compatibility, Experimental, and Internal surfaces are explicit.
• Adds Experimental application wrappers for spherical features, relative pose, CAM-guided localization, Mask R-CNN workflow composition, classical stereo, and registered full-sphere dense stereo.

Dense stereo, relative pose, DoG application wrappers, CAM/SAM workflows, graph, reconstruction, and SLAM remain Experimental. This release does not include datasets, checkpoints, model weights, training pipelines, or private evidence.

Install with:

pip install panorai==3.8.0

See the full changelog, the 3.8.0 release checklist, and changes since 3.7.0.

PanorAi 3.7.0

PanorAi 3.7.0 is a transition release that consolidates the library as a product while preserving the 3.x compatibility surface and making the 4.0 direction explicit.

Highlights

• Adds the Experimental panorai-spatial-semantic-graph/v1 under panorai.graph, including typed evidence, pair association/localization, spatial posteriors, deterministic graph lifecycle, portable JSONL/NPZ persistence, replay, and evidence-aware queries.
• Introduces a JSON-Schema-validated public API inventory with domain, product family, maturity, visibility, canonical import, replacement, and promotion gates.
• Keeps existing 3.x Stable and Compatibility imports valid. panorai.object_localization remains a transition facade over the new graph association/localization implementation.
• Reorganizes the documentation around product capabilities, concepts, reference, evidence, development, releases, and the published 4.0 migration map.
• Adds benchmark cataloging and public citation, contribution, governance, security, and support policies.

No existing Experimental API is promoted to Stable in this release. panorai.graph is Experimental and its builder consumes explicit perception results; it does not run detection, matching, neural inference, SLAM, or reconstruction.

Install with:

pip install panorai==3.7.0

See the full changelog and changes since 3.6.0.

PanorAi 3.6.0

PanorAi 3.6.0 adds opt-in experimental spherical deep-learning portability and a checksum-pinned external metric-depth adapter, refreshes the PyPI project summary, and removes legacy vendored research trees and stale scaffolding. Model sources and weights remain external, opt-in, and never redistributed by PanorAi. See CHANGELOG.md for the complete contract and migration details.

PanorAi 3.5.0

What's Changed

• features: add relative-pose reference profile by @rlsgarcia-code in #173
• features: add experimental resolution selection report by @rlsgarcia-code in #174
• Make scanner shadow padding state explicit by @RobinsonGarcia in #175
• Add experimental spherical coarse-to-fine detector by @rlsgarcia-code in #176
• feat: promote optimized spherical feature frontend by @rlsgarcia-code in #177
• docs: map spherical workflows and benchmarks by @RobinsonGarcia in #178
• chore: prepare PanorAi 3.5.0 release by @rlsgarcia-code in #179

Full Changelog: v3.4.0...v3.5.0

PanorAi 3.4.0

PanorAi 3.4.0

PanorAi 3.4.0 adds direct spherical dense stereo as an explicitly
Experimental API, together with an OpenCV-style tutorial, a detailed
geometry and optimization article, visualization helpers, and reproducible
real-texture documentation assets.

Highlights

• New panorai.stereo package for pose-constrained ERP-to-ERP radial-range
  estimation from a known metric relative pose.
• Adaptive coarse-to-fine inverse-range search, seam-aware sampling,
  four-path SGM-like aggregation, bidirectional consistency, confidence and
  validity diagnostics.
• First-party native spherical-filter acceleration with NumPy fallback.
• colorize_spherical_range and render_spherical_stereo_result helpers.
• Expanded two-view geometry documentation, including known camera-height
  metric reasoning and rendered SVG flowcharts.
• Full CPython 3.11–3.14 wheel matrix for Linux x86_64/aarch64, macOS
  x86_64/arm64, and Windows AMD64.

Stability boundary

The dense-stereo interface is
panorai-spherical-dense-stereo/v1-experimental. It requires caller-supplied
calibrated R,t; its output is radial range in the translation's units, not
camera-axis z-depth. Current evidence supports experimental use and evaluation,
not a Stable accuracy guarantee. Existing Stable geometry-v1 and projection
contracts are unchanged.

See CHANGELOG.md and the 3.4.0 documentation for the complete details and
known limitations.

PanorAi 3.3.1

Patch release restoring the cross-platform publication gate. Geometry-v1 checksum fixtures are now forced to LF on every checkout, with an independent hosted Windows integrity gate and a shared raw-byte verifier. No runtime API, numerical, dependency, or native-kernel behavior changed from the reviewed 3.3.0 candidate. The 3.3.0 workflow stopped before TestPyPI, so no 3.3.0 package files were published.

PanorAi 3.3.0

PanorAi 3.3.0 completes the stable spherical projection and feature-extraction workflow while keeping pairwise pose, PyCOLMAP export, multiview reconstruction, multiscale routing, and SLAM explicitly Experimental.

Highlights:

• Stable panorai-object-workflow/v1 for modality-aware arbitrary-N view generation and reconstruction.
• Stable panorai-spherical-features/v1 extraction/matching core with OpenCV-backed SIFT, ORB, AKAZE, BF, and FLANN presets.
• First-party C++17 acceleration for compatible NumPy cubemap and arbitrary-N projection/reconstruction paths, plus native Essential and bundle-adjustment kernels.
• Twenty native wheels covering CPython 3.11–3.14 on Linux x86_64/aarch64, macOS x86_64/arm64, and Windows AMD64.
• A redesigned, use-case-first README and an educational documentation path from spherical pixels and rays through samplers/blenders, feature matching, spherical Essential geometry, triangulation, and multiview reconstruction.
• Reproducible tutorial figures derived from a checksum-pinned public CC0 Poly Haven panorama; wheels remain free of raster documentation media.

The release workflow validates the full source suite, Torch parity/autograd, dependency boundaries, native kernels, all twenty installed wheels, strict documentation, artifact/license policy, TestPyPI identity, production PyPI identity, provenance, and Pages from the same immutable tag.

See the full changelog: https://github.com/rlsgarcia-code/PanorAi/blob/v3.3.0/CHANGELOG.md

PanorAi 3.2.0

PanorAi 3.2.0 adds an Experimental ergonomic workflow over the stable panorai.geometry foundation while preserving the 3.x API.

Highlights

• Start from panorai.EquirectangularImage and attach radial depth or integer labels with explicit validity.
• Generate deterministic cube, Fibonacci, icosahedron, or spiral views without manually assembling channels or projection plans.
• Apply array or tensor models through views.map() and reconstruct with modality-safe interpolation and blending.
• Use process_views() as the exact shorthand for views().map().reconstruct().
• Inspect the executed geometry, modalities, interpolation, validity, backend, device, and reconstruction policy with views.describe().
• NumPy HW/HWC and Torch HW/CHW/NCHW workflows preserve backend, batch, dtype, device, support, validity, and autograd.

Semantics and compatibility

The new surface is Experimental. Existing 3.x names and advanced APIs remain available. Image and radial depth are continuous and bilinear; labels are nearest-only. Strict invalid-data propagation remains the default, with explicit mask-authoritative renormalization available as an opt-in policy.

The distribution retains lightweight adapters for optional depth backends but does not vendor model implementations, datasets, checkpoints, or training trees.

The protected release workflow builds one wheel and one sdist, audits and attests them, verifies byte identity through TestPyPI, runs installed NumPy/Torch geometry and documentation checks, and promotes the same files to PyPI only after protected approval.

Full details: https://github.com/rlsgarcia-code/PanorAi/blob/v3.2.0/CHANGELOG.md

PanorAi 3.1.0

PanorAi 3.1.0 establishes panorai.geometry as the stable, convention-explicit API for spherical image processing and projection.

Highlights

• Stable ERP pixel/ray, rectangular gnomonic, and canonical cubemap geometry.
• Immutable NumPy projectors with optional Torch backends and input autograd.
• Explicit geometric support, data validity, and radial-depth semantics.
• Safe modality rules: labels and masks use nearest sampling; bilinear continuous data supports strict propagation by default and explicit mask-authoritative renormalization.
• Real projector/blender integration with mask-correct handling of black, zero-valued, unsupported, and non-finite samples.
• Python 3.10–3.12 release verification, installed-wheel conformance, independent geometry oracles, strict documentation, reproducible artifacts, provenance attestations, and Trusted Publishing.

Compatibility and distribution

The recorded 3.0 public names remain available. Corrected behavior and migration guidance are documented in the changelog and geometry specification.

The distribution contains lightweight adapters for optional depth backends, not vendored model implementations, datasets, checkpoints, or training code. Install and review each upstream backend under its own requirements and license.

The release workflow builds one wheel and one sdist, validates the exact files through TestPyPI, and promotes those same files to PyPI after the protected production approval.

Full details: https://github.com/rlsgarcia-code/PanorAi/blob/v3.1.0/CHANGELOG.md

v3.0.19

Update setup.py