amd/mobilenet_v2_1.0_224

MobileNetV2

MobileNetV2 is an image classification model pre-trained on ImageNet-1k dataset at resolution 224x224. It was introduced in the paper MobileNetV2: Inverted Residuals and Linear Bottlenecks by Mark Sandler et al. and first released in this repository.

We develop a modified version that could be supported by AMD Ryzen AI.

Model description

MobileNetV2 is a simple network architecture that allows to build a family of highly efficient mobile models. It allows memory-efficient inference. MobileNetV2 is a model typically used for image classification tasks. And also can be used for object detection and image segmentation tasks. All tasks show competitive results.

The model is named mobilenet_v2_depth_size, for example, mobilenet_v2_1.4_224, where 1.4 is the depth multiplier and 224 is the resolution of the input images the model was trained on.

Intended uses & limitations

You can use the raw model for image classification. See the model hub to look for fine-tuned versions on a task that interests you.

How to use

Installation

1. Follow Ryzen AI Installation to prepare the environment for Ryzen AI.

2. Run the following script to install pre-requisites for this model.

   pip install -r requirements.txt 
   

Test & Evaluation

• Inference one image (Image Classification):

  import sys
  import onnxruntime
  import torch
  import torchvision.transforms as transforms 
  from PIL import Image
  
  image_path = sys.argv[1]
  onnx_model = sys.argv[2]
  
  normalize = transforms.Normalize(
    mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
  img_transformer = transforms.Compose([
                  transforms.Resize(256),
                  transforms.CenterCrop(224),
                  transforms.ToTensor(),  
                  normalize])
  img_tensor = img_transformer(Image.open(image_path)).unsqueeze(0)
  img_tensor = torch.permute(img_tensor, (0, 2, 3, 1))
  so = onnxruntime.SessionOptions()
  ort_session = onnxruntime.InferenceSession(
          onnx_model, so, 
          providers=['CPUExecutionProvider'], 
          provider_options=None)
  input = img_tensor.numpy()
  ort_input = {ort_session.get_inputs()[0].name: input}
  
  output = ort_session.run(None, ort_input)
  top5_probabilities, top5_class_indices = torch.topk(torch.nn.functional.softmax(torch.tensor(output[0])), k=5)
  

• Evaluate ImageNet validation dataset (50,000 Images), using eval_onnx.py .

  • Test accuracy of the quantized model on CPU.

    python eval_onnx.py  --onnx_model=./mobilenetv2_int8.onnx --data_dir=./{DATA_PATH}
    

  • Test accuracy of the quantized model on IPU.

    python eval_onnx.py  --onnx_model=./mobilenetv2_int8.onnx --data_dir=./{DATA_PATH} --ipu --provider_config Path\To\vaip_config.json
    

  • Users can use vaip_config.json in folder voe-4.0-win_amd64 of ryzen-ai-sw-1.0.zip file.

​ DATA_PATH: Path to ImageNet dataset where contains the validation folder.

Performance

Dataset: ImageNet validation dataset (50,000 images).

Metric	Accuracy on IPU
top1& top5 accuracy	75.62% / 92.52%

Citation

@article{MobileNet v2,
  author       = {Mark Sandler and
                  Andrew G. Howard and
                  Menglong Zhu and
                  Andrey Zhmoginov and
                  Liang{-}Chieh Chen},
  title        = {MobileNetV2: Inverted Residuals and Linear Bottlenecks},
  year         = {2018},
  url          = {http://arxiv.org/abs/1801.04381},
}