推論処理システム、サーバ、エッジデバイス、及びこれらの制御方法、並びにプログラム

Abstract

To provide an inference processing system, a server, an edge device, control methods thereof, and programs that can shorten the total calculation processing time while keeping a security risk low.SOLUTION: An inference processing system 1 performs calculation of an inference model using a server 103 and an imaging device 101. In the server 103, a first estimation unit 303 sets a server-side inference model that is the same as an inference model. In the imaging device 101, a second estimation unit 313 sets an edge-side inference model consisting of an input layer and first to second intermediate layers of the inference model, and an image identification unit 220 generates layer position information according to the size of a human face in an input image. When calculation results (layer data) obtained by processing the input image from the input layer to a layer indicated by the layer position information using the edge-side inference model, and the layer position information are transmitted from the imaging device 101 to the server 103, in the server-side inference model, the layer data is input to the layer next to the layer indicated by the layer position information and processed up to an output layer.SELECTED DRAWING: Figure 4

Images