IMAGING APPARATUS, INFORMATION PROCESSING APPARATUS, AND METHOD FOR CONTROLLING THESE, AND PROGRAM AND LEARNED MODEL SELECTION SYSTEM

To provide an imaging apparatus that can replace as appropriate a learned model for performing processing related to a picked-up image, an information processing apparatus, and a method for controlling these, and a program and a learned model selection system.SOLUTION: An imaging apparatus C communi...

Full description

Saved in:
Bibliographic Details
Main Author TANABE AKIHIRO
Format Patent
LanguageEnglish
Japanese
Published 17.06.2021
Subjects
ACCESSORIES THEREFOR
APPARATUS OR ARRANGEMENTS EMPLOYING ANALOGOUS TECHNIQUES USINGWAVES OTHER THAN OPTICAL WAVES
APPARATUS OR ARRANGEMENTS FOR TAKING PHOTOGRAPHS OR FORPROJECTING OR VIEWING THEM
CALCULATING
CINEMATOGRAPHY
COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
COMPUTING
COUNTING
ELECTRIC COMMUNICATION TECHNIQUE
ELECTRICITY
ELECTROGRAPHY
HOLOGRAPHY
PHOTOGRAPHY
PHYSICS
PICTORIAL COMMUNICATION, e.g. TELEVISION
Online AccessGet full text

Cover

More Information
Summary:To provide an imaging apparatus that can replace as appropriate a learned model for performing processing related to a picked-up image, an information processing apparatus, and a method for controlling these, and a program and a learned model selection system.SOLUTION: An imaging apparatus C communicates with a server SV that holds a learned model corresponding to a plurality of classes, and executes object detection during AF processing by using a learned model corresponding to a subject preferential in a neural network processing unit 205. A CPU 201 of the imaging apparatus C generates history data based on a picked-up image, updates the preferential subject based on the history data, acquires, from the server SV, a learned coefficient parameter of the learned model corresponding to the updated preferential subject, and based on the acquired learned coefficient parameter, changes the setting of the learned model in the neural network processing unit 205.SELECTED DRAWING: Figure 6 【課題】撮像画像に関する処理を行う学習済みモデルを適宜入れ替えることができる撮像装置、情報処理装置、及びこれらの制御方法並びにプログラム、学習済みモデル選択システムを提供する。【解決手段】複数のクラスの夫々に対応する学習済みモデルを保持するサーバＳＶと通信する撮像装置Ｃは、ニューラルネットワーク処理部２０５で優先する被写体に対応する学習済みモデルを用いてＡＦ処理の際の物体検出を実行する。撮像装置ＣのＣＰＵ２０１は撮像画像に基づき履歴データを作成し、その履歴データに基づき優先する被写体を更新し、その更新した優先する被写体に対応する学習済みモデルの学習済み係数パラメータをサーバＳＶから取得し、その取得した学習済み係数パラメータに基づいてニューラルネットワーク処理部２０５における学習済みモデルの設定を変更する。【選択図】図６
Bibliography:Application Number: JP20190221408