Dynamic Shape with Binary Inputs

Introduction

This guide shows how to use the binary inputs feature to send data in binary format. This means you can load just a JPEG or PNG image and run inference on it without any data preprocessing.

To run inference on binary encoded data, make sure your model accepts the NHWC layout. When preparing the request, we need to let the Model Server know that the data is in binary format.

Learn more about the binary inputs feature.

Steps

Download the Pretrained Model

Download the model files and store them in the models directory

mkdir -p models/resnet/1
curl https://storage.openvinotoolkit.org/repositories/open_model_zoo/2022.1/models_bin/2/resnet50-binary-0001/FP32-INT1/resnet50-binary-0001.bin https://storage.openvinotoolkit.org/repositories/open_model_zoo/2022.1/models_bin/2/resnet50-binary-0001/FP32-INT1/resnet50-binary-0001.xml -o models/resnet/1/resnet50-binary-0001.bin -o models/resnet/1/resnet50-binary-0001.xml

Pull the Latest Model Server Image from Docker Hub

Pull the latest version of OpenVINO™ Model Server from Docker Hub :

docker pull openvino/model_server:latest

Start the Container with Downloaded Model

Start the container with the image pulled in the previous step and mount the models directory :

docker run --rm -d -v $(pwd)/models:/models -p 9000:9000 openvino/model_server:latest --model_name resnet --model_path /models/resnet --layout NHWC:NCHW --port 9000

Download ovmsclient Package

pip3 install ovmsclient 

Download a Sample Image and Label Mappings

wget https://raw.githubusercontent.com/openvinotoolkit/model_server/releases/2024/0/demos/common/static/images/zebra.jpeg

wget https://raw.githubusercontent.com/openvinotoolkit/model_server/releases/2024/0/demos/common/python/classes.py

Run Inference

echo '
import numpy as np 
from classes import imagenet_classes 
from ovmsclient import make_grpc_client 
client = make_grpc_client("localhost:9000") 

with open("zebra.jpeg", "rb") as f: 
    img = f.read() 

output = client.predict(inputs={ "0": img}, model_name= "resnet") 
result_index = np.argmax(output[0]) 
print(imagenet_classes[result_index])' >> predict.py

python3 predict.py
zebra