LEARNING METHOD, INFORMATION PROCESSING DEVICE AND LEARNING PROGRAM

PROBLEM TO BE SOLVED: To improve classification accuracy.SOLUTION: A calculation part 1b acquires an image G1, information indicating a first area G11 in the image G1, and information indicating a classification destination of the image G1. The calculation part 1b compares a first detection result of plural kinds of feature amounts in the first area G11, and a second detection result of plural kinds of feature amounts in a second area G12 other than the first area G11, in the image G1. The calculation part 1b determines a weighting of each of the plural kinds of feature amounts, when the image G1 is classified to the acquired classification destination according to the comparison. The calculation part 1b performs learning of classification of the image G1 based on a weighting result of the plural kinds of feature amount by weighting.SELECTED DRAWING: Figure 1 【課題】分類精度を上げること。【解決手段】演算部１ｂは、画像Ｇ１と画像Ｇ１のうち第１領域Ｇ１１を示す情報と画像Ｇ１の分類先を示す情報とを取得する。演算部１ｂは、第１領域Ｇ１１における複数の種類の特徴量の第１検出結果、および、画像Ｇ１のうち第１領域Ｇ１１以外の第２領域Ｇ１２における複数の種類の特徴量の第２検出結果を比較する。演算部１ｂは、当該比較に応じて、取得した分類先へ画像Ｇ１を分類する際の複数の種類の特徴量それぞれの重みを決定する。演算部１ｂは、複数の種類の特徴量それぞれを当該重みにより重み付けした結果に基づいて、画像Ｇ１の分類の学習を行う。【選択図】図１