lamina op• 資料種類:matrix → matrix
• 呼叫: import fullseye as fs; fs.ledger.fly_onoff_split(movie, dt_s, tau_on_s=0.02, tau_off_s=0.02, rectify=True)(要直接呼叫實作,import flyvision; flyvision.fly_onoff_split(movie, dt_s, tau_on_s=0.02, tau_off_s=0.02, rectify=True);從台帳取用則 opsflyvision.get("fly_onoff_split"))
> 該運算子的說明尚無譯文,以下照原文給出。
Split a contrast movie into the ON and OFF channels the medulla carries.
Beyond the lamina the fly stops carrying one signed signal and carries two:
an ON channel (Mi1 / Tm3, brightening) and an OFF channel (Tm1 / Tm2,
darkening), each with its own relay dynamics. This op is that split, and it
keeps the rectification *optional* on purpose: the split was measured
downstream, in the motion response (Joesch et al., *Nature* 468:300, 2010),
while L1/L2 themselves respond linearly (Clark et al., *Neuron* 70:1165,
2011), so a pathway that rectifies at the lamina is making a claim the
recordings do not.
movie: `(T, n)` contrast, rows = time (the return of
:func:fly_lamina_filter). dt_s: sample interval, seconds.
tau_on_s / tau_off_s: the low-pass time constant of each channel, seconds.
rectify: `True half-wave rectifies (ON = max(c, 0)`,
`OFF = max(-c, 0), both >= 0); False` passes the signed contrast into
the ON channel and its negation into the OFF channel, which is the linear
L1/L2 case and lets the rectification happen downstream instead.
Returns `(T, 2n) float64: columns 0..n-1 are ON, n..2n-1` are OFF,
the same ommatidium order in each half.
Ground truth (exact, with `tau_on_s == tau_off_s` so the two channels share
one filter):
• `ON - OFF` is the low-passed input, whichever *rectify* you chose;
• with `rectify=True, ON + OFF` is the low-passed absolute value;
• a movie that never goes negative leaves the OFF channel identically zero
(and vice versa) — the split does not invent a dark event.
Raises `ValueError`: a non-2-D / too-short / non-finite *movie*, a movie
over :data:MAX_MOVIE_ELEMENTS, and a non-positive *dt_s* / *tau_on_s* /
*tau_off_s*.
• 範例資料目錄(下載 URL / 授權) —— 2-D 用 skimage.data(BSD/公有領域)加合成圖,3-D 給出真實資料源(Stanford/PDS 等)的下載 URL。
• 運算子來歷與參考文獻 —— 該運算子族所依據的研究/方法出處。
• 演算法的正典(作者・年份)與用途見上面的族使用指南。
• poc_fly_optomotor_steering — py -3.11 examples/poc_fly_optomotor_steering.py
matrix 作為輸入)fly_lamina_filter · fly_t4t5_field · fly_flow_from_directions · fly_egomotion_from_flow · fly_hs_readout
lamina)*Provenance: flyvision.py — FLYVISION 運算子登記表。本條目由 tools/opdocs.py md 自動產生(請勿手動編輯)。*
© 2026 Kazufumi Furuse — Fullseye operator documentation. Licensed under Apache-2.0.