Spatio-temporal coding in the retinal circuit

The retina is more sophisticated than many physiologists and neuromorphic engineers think. In the retina, image of real world is first continuously transduced to spatio-temporal distribution graded responses (slow analog voltage changes). In ganglion cells (output neurons of the retinal circuits), the graded potentials are transformed to spikes following some dynamical computations by neuronal circuits. In this regard, image of real world is considered to be encoded with spatio-temporal distribution of spikes. Recent physiological studies have demonstrated several lines of evidence suggesting that not only the spike rate but also the spike timing among ganglion cells play roles to encode the image. We developed a retino-morphic vision sensor (RMVS) consisting of analog silicon retina and FPGA. The RMVS mimics layered structure of the retinal circuits with a Linear, Nonlinear and spike-generation modules. Taking advantages the analog and digital technologies, the RMVS can reproduce the event-based spike responses of different types of ganglion cell in real time. We propose to study together how natural images are encoded with such simple circuit structure using the RMVS and seek possible applications of the coding strategies of the retinal circuits to the robotic vision.

Go to group wiki

Timetable

Day Time Location
Wed, 25.04.2018 16:00 - 17:00 Disco
Thu, 26.04.2018 19:00 - 20:00 Disco
Fri, 27.04.2018 14:00 - 15:00 Disco
Mon, 30.04.2018 16:00 - 17:00 Disco
Tue, 01.05.2018 11:00 - 12:00 Disco
Wed, 02.05.2018 16:00 - 17:00 Disco
Thu, 03.05.2018 11:00 - 12:00 Disco

Moderator

Naofumi Suematsu
Tetsuya Yagi

Members

Evripidis Gkanias
jacques kaiser
Jamie Knight
Gregor Lenz
Shih-Chii Liu
Emre Neftci
Omar Oubari
Nicoletta Risi
Baris Serhan
Benjamin Ward-Cherrier
Tetsuya Yagi
Ali Zeinolabedin