LEARNING DATA GENERATION DEVICE, LEARNING DATA GENERATION METHOD, MACHINE LEARNING METHOD, AND PROGRAM
To provide a technology for simplifying a process of constructing learning data by performing annotation with use of CG data.SOLUTION: The learning data generation device includes a first storage unit which stores scene data of CG data including a plurality of models, a first image generation unit w...
Saved in:
Main Authors | , , |
---|---|
Format | Patent |
Language | English Japanese |
Published |
14.02.2019
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | To provide a technology for simplifying a process of constructing learning data by performing annotation with use of CG data.SOLUTION: The learning data generation device includes a first storage unit which stores scene data of CG data including a plurality of models, a first image generation unit which generates a scene data image by using the scene data, a model setting unit which generates, from the scene data, scene data for generating learning data, by deleting any model other than a specific object model from the scene data and making the specific object model in a specific color, a second image generation unit which generates an image for generating learning data by using the scene data for generating learning data, an object region setting unit which sets a region for the specific object model in the image for generating learning data, and a second storage unit which stores, as learning data, the scene data image and the region information in association with each other.SELECTED DRAWING: Figure 1
【課題】CGデータを用いてアノテーション処理し学習データを構築する処理を簡易化する技術を提供する。【解決手段】学習データ生成装置は、複数のモデルを含むCGデータのシーンデータを格納する第一記憶部と、前記シーンデータを用いて、シーンデータ用画像を生成する第一画像生成部と、特定物体モデル以外のモデルを前記シーンデータから削除し、前記特定物体モデルを特定色にするよう、前記シーンデータから学習データ生成用シーンデータを生成するモデル設定部と、前記学習データ生成用シーンデータを用いて、学習データ用画像を生成する第二画像生成部と、前記学習データ用画像において、前記特定物体モデルの領域を設定する物体領域設定部と、前記シーンデータ用画像と前記領域の情報とを対応付けて、学習データとして格納する第二記憶部とを備える。【選択図】図1 |
---|---|
Bibliography: | Application Number: JP20180101636 |