model = dict( type='ImageClassifier', backbone=dict( type='ViTEVA02', arch='b', img_size=224, patch_size=14, sub_ln=True, final_norm=False, out_type='avg_featmap'), neck=None, head=None, ) data_preprocessor = dict( # RGB format normalization parameters mean=[0.48145466 * 255, 0.4578275 * 255, 0.40821073 * 255], std=[0.26862954 * 255, 0.26130258 * 255, 0.27577711 * 255], # convert image from BGR to RGB to_rgb=True, )