onnx/ShuffleNet_V1

1import onnx2import numpy as np3import onnxruntime as ort4from PIL import Image5import cv26import os7import gradio as gr8 9import mxnet10from mxnet.gluon.data.vision import transforms11 12os.system("wget https://s3.amazonaws.com/onnx-model-zoo/synset.txt")13 14 15with open('synset.txt', 'r') as f:16    labels = [l.rstrip() for l in f]17    18os.system("wget https://github.com/AK391/models/raw/main/vision/classification/shufflenet/model/shufflenet-9.onnx")19 20os.system("wget https://s3.amazonaws.com/model-server/inputs/kitten.jpg")21 22 23 24model_path = 'shufflenet-9.onnx'25model = onnx.load(model_path)26session = ort.InferenceSession(model.SerializeToString())27 28def get_image(path):29    with Image.open(path) as img:30        img = np.array(img.convert('RGB'))31    return img32    33    34def preprocess(img):35    '''36    Preprocessing required on the images for inference with mxnet gluon37    The function takes loaded image and returns processed tensor38    '''39    img = np.array(Image.fromarray(img).resize((224, 224))).astype(np.float32)40    img[:, :, 0] -= 123.6841    img[:, :, 1] -= 116.77942    img[:, :, 2] -= 103.93943    img[:,:,[0,1,2]] = img[:,:,[2,1,0]]44    img = img.transpose((2, 0, 1))45    img = np.expand_dims(img, axis=0)46 47    return img48 49def predict(path):50    img = get_image(path)51    img = preprocess(img)52    ort_inputs = {session.get_inputs()[0].name: img}53    preds = session.run(None, ort_inputs)[0]54    preds = np.squeeze(preds)55    a = np.argsort(preds)56    results = {}57    for i in a[0:5]:    58        results[labels[a[i]]] = float(preds[a[i]])59    return results60       61 62title="ShuffleNet-v1"63description="ShuffleNet is a deep convolutional network for image classification. ShuffleNetV2 is an improved architecture that is the state-of-the-art in terms of speed and accuracy tradeoff used for image classification."64 65examples=[['kitten.jpg']]66gr.Interface(predict,gr.inputs.Image(type='filepath'),"label",title=title,description=description,examples=examples).launch(enable_queue=True,debug=True)