IBM 视觉识别分类器状态失败

Question

我有以下 IBM Watson Visual Recognition Python SDK 用于创建简单的分类器：

with open(os.path.dirname("/home/xxx/Desktop/Husky.zip/"), 'rb') as dogs, \ 
    open(os.path.dirname("/home/xxx/Desktop/Husky.zip/"), 'rb') as cats:
    print(json.dumps(visual_recognition.create_classifier('Dogs Vs Cats',dogs_positive_examples=dogs,negative_examples=cats), indent=2))

带有新分类器 ID 及其状态的响应如下：

{
  "status": "training", 
  "name": "Dogs Vs Cats", 
  "created": "2016-06-23T06:30:00.115Z", 
  "classes": [
    {
      "class": "dogs"
    }
  ], 
  "owner": "840ad7db-1e17-47bd-9961-fc43f35d2ad0", 
  "classifier_id": "DogsVsCats_250748237"
}

训练状态显示失败。

打印(json.dumps(visual_recognition.list_classifiers(), 缩进=4))

{
    "classifiers": [
        {
            "status": "failed", 
            "classifier_id": "DogsVsCats_250748237", 
            "name": "Dogs Vs Cats"
        }
    ]
}

这是什么原因？

Answer 1

训练调用和数据有大小限制：

The service accepts a maximum of 10,000 images or 100 MB per .zip file

The service requires a minimum of 10 images per .zip file.

The service accepts a maximum of 256 MB per training call.

分类调用也有大小限制：

The POST /v3/classify methods accept a maximum of 20 images per batch.

The POST /v3/detect_faces methods accept a maximum of 15 images per batch.

The POST /v3/recognize_text methods accept a maximum of 10 images per batch.

见http://www.ibm.com/smarterplanet/us/en/ibmwatson/developercloud/doc/visual-recognition/customizing.shtml

Answer 2

with open(os.path.dirname("/home/xxx/Desktop/Husky.zip/"), 'rb') as dogs, \ 
    open(os.path.dirname("/home/xxx/Desktop/Husky.zip/"), 'rb') as cats:
    print(json.dumps(visual_recognition.create_classifier('Dogs Vs Cats',dogs_positive_examples=dogs,negative_examples=cats), indent=2))

您正在发送相同的文件内容 "Husky.zip" 供服务用作正例和反例。然而，系统需要至少 10 个正例图像和 10 个负例图像，它们是 独特的 。该服务在训练前比较图像文件内容的哈希码，并只在正集中留下任何重复项。因此，去重后你的否定集是空的，导致训练失败。在您的分类器详细信息的详细列表中应该有一个名为 "explanation" 的附加字段，说明这可能是问题所在。

IBM 视觉识别分类器状态失败

IBM Visual Recognition Classifier status failed

python-2.7

ibm-cloud

ibm-watson

visual-recognition