fly_eye_merge — FLYVISION selfmotion op

• Data kinds: table × table → table

• Call: import fullseye as fs; fs.ledger.fly_eye_merge(lattice_a, lattice_b, *more) (to call the implementation directly, import flyvision; flyvision.fly_eye_merge(lattice_a, lattice_b, *more); from the registry, opsflyvision.get("fly_eye_merge"))

Usage

Several lattices seen as one wide eye — the viewing directions of all of them.

:func:fly_hex_resample needs a lattice that fits inside one pinhole image,

which caps a single patch at well under a hemisphere. A fly is not so

limited: its two compound eyes together see almost the whole sphere, and the

wide-field cells that read self-motion out of them integrate over all of it.

This op is that integration made explicit — render each patch through its own

camera, run the pathway on each, then merge the *geometry* so that one

least-squares fit sees every ommatidium at once.

It matters more than it looks. In a naturalistic 1/f scene the response of a

correlation detector is contrast-weighted, so a narrow patch is at the mercy

of whichever few large features happen to be in it; widening the field is

what turns the estimate from a guess into a measurement (the PoC measures how

much).

lattice_a / lattice_b / *more: two or more :func:fly_hex_lattice results. They must

share the inter-ommatidial angle and the geometry (different spacings are

different eyes and the merged field would silently mix two sampling scales),

and no two ommatidia may look in exactly the same direction (merging a patch

with itself would double its vote without saying so).

Returns a lattice dict with the same keys plus `"eye"`, the index of the

patch each ommatidium came from, in input order. The merged lattice is what

:func:fly_matched_filter and :func:fly_egomotion_from_flow take;

:func:fly_hex_resample and :func:fly_flow_from_directions stay per patch,

because a pinhole image and a hexagonal neighbourhood are both local.

Raises `ValueError`: a malformed lattice (including a sequence handed

in where a lattice was expected), a mismatched `dphi_rad or geometry`,

and a direction that appears twice.

Detailed usage guide

• fly_vision family guide

References (sample data, literature)

• Sample-data catalog (download URLs / licences) — 2-D uses skimage.data (BSD/public domain) plus synthetic images; 3-D lists download URLs for real data sources (Stanford, PDS, …).

• Operator provenance and references — the sources of the research/methods this op family came from.

• The canonical algorithm (author, year) and its uses are named in the family usage guide above.

Runnable examples (verified samples that actually call this op)

• poc_fly_optomotor_steering — py -3.11 examples/poc_fly_optomotor_steering.py

Ops the type connects to (they accept table as input)

fly_t4t5_field · fly_flow_from_directions · fly_matched_filter · fly_egomotion_from_flow · fly_hex_resample · fly_hs_readout

Same category (selfmotion)

fly_matched_filter · fly_egomotion_from_flow


*Provenance: flyvision.py — FLYVISION operator registry. This per-op note is generated by tools/opdocs.py md (do not hand-edit).*

© 2026 Kazufumi Furuse — Fullseye operator documentation. Licensed under Apache-2.0.