The template provides 10 images for mules, horses, and alpacas, respectively. You can use these images to train a model in the Auto Learning platform and then use the model to classify mules, horses, and alpacas with an accuracy of higher than 80%.
Deep learning is frequently used in image classification. Image classification adds labels to a group of images to classify the images. For example, if image A contains characters, then the image is classified as character images. If image B contains flowers or plants, then the image is classified as natural images. When an image is input into an image classification model, the model automatically returns the class of the image to you. Image classification is used in multiple scenarios, such as photo classification, animal classification, and plant classification.
The Auto Learning platform of Alibaba Cloud Machine Learning Platform for AI is a high-performance platform based on transfer learning. It has no requirements on coding skills and the amount of data used to train models. In addition, it supports hyperparameter tuning and optimization. With the Auto Learning platform, you can train models extremely fast. The Auto Learning platform does not require any algorithm knowledge. You only need to provide a small amount of training samples. This topic uses the mule, horse, and alpaca classification model as an example to describe how to train a model in the Auto Learning platform in three steps.
Move your mouse to the template, and then click Create from Template to create an instance. After the instance is created, it is added to the instances list. Use the provided mule, horse, and alpaca images as training data.Note: This topic describes how to create an instance from a template and uses the images provided by the Auto Learning platform to train the model. Therefore, you do not need to activate and authorize Object Storage Service (OSS).
After the instance is created from the template, you are navigated to the data labeling page of the instance. On the data labeling page, you can add labels to images. After you click Start Training, the platform shows the image labeling and classification results. You can customize the maximum model training duration. If the maximum model training duration is set to 20 minutes, when the model training duration reaches 20 minutes, the instance stops model training and returns the trained model.
- Image labeling page
- Set the maximum model training duration
You can view the entire training procedure on the model training page, including the training progress and time consumption. You can also manage model versions. After model training is complete, statistics about the trained model, such as the accuracy and precision, are displayed. If you want to increase the accuracy of the model, you can input more training data or extend the maximum training duration.
The platform provides the model testing function for you to test the generated image classification model. To test the model, you only need to upload an image. The platform then returns the image classification result. As shown in the following figure, the mule, horse, and alpaca classification model returns the expected result.
By following the preceding steps, now you have a mule, horse, and alpaca classification model. You can click Go to EAS to deploy the trained model as an online service.