Developing deep learning based regression approaches for prediction of firmness and pH in Kyoho grape using Vis/NIR hyperspectral imaging

•Grape firmness and pH were successfully predicted by HSI and deep learning method.•CARS and SPA were studied and compared to obtain conventional features for prediction.•Deep spectral features in Vis/NIR hyperspectral image of grape were extracted by SAE.•Proposed deep learning models were used to fit the deep spectral features to firmness and pH.