STARLESS
Pure Stellar Segregation
Unveiling the latent universe.
Starless is a native, hardware-accelerated neural network architecture engineered to isolate and segregate stellar profiles from deep-sky imaging data. By operating directly on raw 32-bit linear FITS data, Starless isolates stellar components to preserve the underlying emission structures as a pure mathematical residual, preventing structural collapse and ensuring absolute scientific fidelity.
Target 01
High-capacity mathematical segregation of complex stellar fields. Drag the handle to compare the original raw frame with the Starless output, highlighting the complete preservation of delicate background structures.
Scaled for
extreme data.
From compact community models to a 73M-parameter core. Zenith and Nadir power lightweight community workflows. Axiom is Starless’ high-capacity architecture. With Axiom 2.1, Starless moves from 13M to 73M parameters, giving the model far more capacity for difficult astrophotography data.
Axiom 2.1
High CapNew Starless core · high-capacity architecture for harder datasets (5.6× larger than Axiom v2)
Axiom v2
PreviousPrevious Starless core · solid baseline architecture
Nadir
DerivedCompact Axiom-derived model · open distribution (Axiom → Nadir)
Zenith
OptimizedAd hoc for Siril · optimized for speed and integration
Target 02
Advanced segregative processing over dense nebular backgrounds. The neural model isolates and extracts stellar sources with sub-pixel mathematical precision, revealing hidden gas structures with high fidelity.
Engineered for pure
data processing.
Signal Separation
With the release of the Axiom 2.1 architecture, signal separation achieves unprecedented mathematical stability. The network isolates stellar profiles down to sub-pixel resolution, minimizing star halos and producing mathematically clean background transitions. This guarantees that faint molecular structures and deep nebula gradients remain intact, allowing for extreme post-processing stretching and unconstrained local editing without artifacts.
Linear Fidelity
Supports unstretched linear-data inputs directly from integration.
High-Density Solvers
Efficient parallelization manages massive stellar aggregates.
Immaculate Star Mask
The generated star layer maps micro-stars and aggressive stellar halos with high spatial consistency. This yields a precise mathematical mask that allows for seamless re-integration, custom stellar reduction, or independent star-field processing without bleed.
Workflow Ready
Designed to integrate directly into advanced astrophotography pipelines. Native support for 32-bit floating-point FITS files, metadata preservation, and accelerated local hardware inference via CoreML and DirectML engines.
Target 03
Precision separation of overlapping signals in high-density galactic centers. Preserves dark dust lanes, faint halos, and complex molecular boundaries without inducing structural degradation.
Acquire the logic.
Deploy Starless Axiom 2.1 in your native environment. Select your build below to gain immediate access to the standalone binaries.
Querying licensing server...