This time we will try to use Baidu's official EasyDL platform to train and deploy a flower recognition network. First, we need to create a model:
You can see that the official has provided a variety of network recognition solutions. Here we choose image classification, create a model, and then import our local dataset:
Here we import the local labeled data set, which includes five kinds of flowers and their corresponding labels. The platform also provides online labeling services, which interested readers can try. Then prepare to train the model:
Here, select Edgeboard (FZ) as the implementation platform , then configure it to high precision and set other configurations to automatic.
Select the training dataset and the corresponding graphics card. The free P4 and 12- core CPU are not bad. Then you can start training:
You can check the option to receive SMS reminders after training is completed, which is a very considerate feature. The training process is a bit long. After the training is completed, you can view the training results of the model:
The EasyDL platform will help us generate a detailed neural network model evaluation report to help us further improve the neural network model. After finalizing the model structure, choose to publish the model:
After the release is successful, the corresponding SDK and test serial number can be generated. The official serial number needs to be paid for use. After obtaining the SDK and serial number, you can copy them into the SD card and use the model normally. The EasyDL platform can build a neural network model simply and quickly, but the user cannot know what network is used to achieve the above functions, trimming and quantization. Therefore, senior developers can optimize the network and develop a neural network inference system with higher performance.
提升卡
变色卡
千斤顶
京公网安备 11010802033920号
