Human-level saccade detection performance using deep neural networks

Bellet, Marie E.; Bellet, Joachim; Nienborg, Hendrikje; Hafed, Ziad M.; Berens, Philipp

doi:10.1101/359018

Cited by 15 publications

(21 citation statements)

References 46 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We analyzed eye movements in all trials. We detected saccades using established methods 41, 75 , and we manually inspected all trials to correct for mis-detections. In experiments requiring a saccade (e.g.…”

Section: Methodsmentioning

confidence: 99%

Perceptual saccadic suppression starts in the retina

Idrees

Baumann

Franke³

et al. 2019

Preprint

Self Cite

View full text Add to dashboard Cite

AbstractVisual sensitivity, probed through perceptual detectability of very brief visual stimuli, is strongly impaired around the time of rapid eye movements. This robust perceptual phenomenon, called saccadic suppression, is frequently attributed to active suppressive signals that are directly derived from eye movement commands. Here we show instead that visual-only mechanisms, activated by saccade-induced image shifts, can account for all perceptual properties of saccadic suppression that we have investigated. Such mechanisms start at, but are not necessarily exclusive to, the very first stage of visual processing in the brain, the retina. Critically, neural suppression originating in the retina outlasts perceptual suppression around the time of saccades, suggesting that extra-retinal movement-related signals, rather than causing suppression, may instead act to shorten it. Our results demonstrate a far-reaching contribution of visual processing mechanisms to perceptual saccadic suppression, starting in the retina, without the need to invoke explicit motor-based suppression commands.

show abstract

Section: Methodsmentioning

confidence: 99%

Perceptual saccadic suppression starts in the retina

Idrees

Baumann

Franke³

et al. 2019

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…We then detected catch-up saccades using velocity and acceleration criteria (Chen and 270 Hafed 2013;Hafed et al 2009;Krauzlis and Miles 1996), and we manually inspected all 271 movements to correct for misses or false detections. For a subset of the data, we used 272 instead a novel state-of-the-art machine-learning approach for saccade detection using 273 convolutional neural networks, which we have recently developed (Bellet et al 2018). 274…”

mentioning

confidence: 99%

The transfer function of the rhesus macaque oculomotor system for small-amplitude slow motion trajectories

Skinner

Buonocore

Hafed

2018

Preprint

Self Cite

View full text Add to dashboard Cite

41Two main types of small eye movements occur during gaze fixation: microsaccades and 42 slow ocular drifts. While microsaccade generation has been relatively well-studied, 43 ocular drift control mechanisms are unknown. Here we explored the degree to which 44 monkey smooth eye movements, on the velocity scale of slow ocular drifts, can be 45 generated systematically. Two male rhesus macaque monkeys tracked a spot moving 46 sinusoidally, but slowly, along the horizontal or vertical directions. Maximum target 47 displacement in the motion trajectory was 30 min arc (0.5 deg), and we varied the 48 temporal frequency of target motion from 0.1 to 5 Hz. We obtained an oculomotor 49 "transfer function" by measuring smooth eye velocity gain (relative to target velocity) as 50 a function of frequency, similar to past work with large-amplitude pursuit. Monkey eye 51 velocities as slow as those observed during slow ocular drifts were clearly target-motion 52 driven. Moreover, like with large-amplitude smooth pursuit, eye velocity gain varied with 53 temporal frequency. However, unlike with large-amplitude pursuit, exhibiting low-pass 54 behavior, small-amplitude motion tracking was band-pass with the best ocular movement 55 gain occurring at ~0.8-1 Hz. When oblique directions were tested, we found that the 56 horizontal component of pursuit gain was larger than the vertical component. Our results 57 provide a catalogue of the control abilities of the monkey oculomotor system for slow 58 target motions, and they also support the notion that smooth fixational ocular drifts are 59controllable. This has implications for neural investigations of drift control and the 60 image-motion consequences of drifts on visual coding in early visual areas.

show abstract

“…Our test battery tests smooth pursuit in a formal way and in this paper we analyze smooth pursuit using a formal model as well. This is indebted due to the current lack of applicable smooth pursuit detection algorithms (Pekkanen and Lappi, 2017; Bellet et al, 2018, but see recent exceptions). We think that the smooth pursuit findings should be treated with caution as our analysis might not generalize to more natural conditions and we had a high number of catch-up saccades.…”

Section: Discussionmentioning

confidence: 99%