3、 Openvino practice: image classification

3、 ... and 、OpenVINO actual combat ： Image classification

Task flow

1. Terminal initialization OpenVINO Environmental Science

   Get into OpenVINO The installation directory , find setupvars.bat File and run , The results are shown in the following figure , That is, the initialization is successful .

   

2. Model acquisition and transformation

   • Call the model downloader and pass –print_all View the models available for download

     

   • adopt –name choice squeezenet1.1 Download the model , The results are shown in the following figure

     • When downloading, if [Error 11004] getaddrinfo failed Tips , The reason lies in host Address by wall , Address raw.githubusercontent.com Add to host In the file .
     • Can be found in –name Add –output_dir Specify download path

     

   • Through the model optimizer （model_optimizer） Convert the model to IR file

     • Get into OpenVINO The model optimizer folder under ./deployment_tools/model_optimizer , Because of the download squeezenet1.1 The model is caffe Model , So run mo_caffe.py File conversion .

       python mo_caffe.py --input_model path --output_dir path

     • After running successfully, get the corresponding bin file and xml file

     

3. Image classification

   • Get ready squeezenet1.1 Label file for squeezenet1.1.labels, The file has been transferred to Baidu cloud
     link ：https://pan.baidu.com/s/17BCKqbTGusnzE5THNz_Spg
     Extraction code ：ysew

     

   • Get a simple classification example from the reasoning engine example

     Copy ./OpenVINO/inference_engine/sample/python/classification_sample_async/ The next path classification_sample_async.py File to your exercise Catalog

   • Run the example , Specify the input image 、 Model 、 Label files and equipment , To get the results
     python classification_samples_async.py -i image_path -m xml_path --labels label_path -d devices_name

     Input the image as ：

     

     The final classification result is ：

     

​