1 about garbage_classify HOT 7 OPEN

你好，那个eval我没有用过呢。val acc和testacc相差不大最多波动会在1%之内。可以考虑不用那个eval，自己直接预测。也可能的原因是train和val是每次都重新划分的，你的val可能不是固定不变的。可以将val重新划分出来，不随训练的变化而变化，希望可以帮助到你发自我的iPhone------------------ 原始邮件 ------------------发件人: Hanpp张 <[email protected]>发送时间: 2019年10月2日 12:06收件人: wusaifei/garbage_classify <[email protected]>抄送: Subscribed <[email protected]>主题: 回复：[wusaifei/garbage_classify]  使用eval 评估训练集 追准确率不高 (#4)我用我自己的数据集 训练时准确率能到 0.7左右 但是 用eval 评估训练集 准确率只有 0.20019050642959。 eval的处理图片函数也没有做归一化 def preprocess_img(img_path, img_size): """ image preprocessing you can add your special preprocess mothod here """ img = Image.open(img_path) resize_scale = img_size / max(img.size[:2]) img = img.resize((int(img.size[0] * resize_scale), int(img.size[1] * resize_scale))) img = img.convert('RGB') img = np.array(img) img = img[:, :, ::-1] img = center_img(img, img_size) return img 我自己加上归一化后也是不高，不知道是什么原因 —You are receiving this because you are subscribed to this thread.Reply to this email directly, view it on GitHub, or mute the thread. [ { "@context": "http://schema.org", "@type": "EmailMessage", "potentialAction": { "@type": "ViewAction", "target": "#4?email_source=notifications\u0026email_token=AGUB6EH5XRWKZNHVIWW6EFDQMQM37A5CNFSM4I4RPQBKYY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4HPAVLAQ", "url": "#4?email_source=notifications\u0026email_token=AGUB6EH5XRWKZNHVIWW6EFDQMQM37A5CNFSM4I4RPQBKYY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4HPAVLAQ", "name": "View Issue" }, "description": "View this Issue on GitHub", "publisher": { "@type": "Organization", "name": "GitHub", "url": "https://github.com" } } ]