Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -1,71 +1,66 @@
|
|
1 |
-
import
|
2 |
-
import math
|
3 |
import matplotlib.pyplot as plt
|
4 |
-
import
|
5 |
-
import
|
6 |
-
import
|
|
|
7 |
import gradio as gr
|
8 |
-
from huggingface_hub import hf_hub_download
|
9 |
-
import onnxruntime as rt
|
10 |
|
11 |
-
|
12 |
-
|
13 |
-
|
14 |
|
15 |
-
|
16 |
-
|
17 |
-
|
18 |
-
|
19 |
-
img =
|
20 |
-
img = np.asarray(img, dtype='float32')
|
21 |
-
# converts jpg pixel value from [0 - 255] to float array [-1.0 - 1.0]
|
22 |
-
img -= [127.0, 127.0, 127.0]
|
23 |
-
img /= [128.0, 128.0, 128.0]
|
24 |
return img
|
25 |
|
26 |
-
#
|
27 |
-
def
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
|
33 |
-
|
34 |
-
|
35 |
-
|
36 |
-
|
37 |
-
img =
|
38 |
-
|
39 |
|
40 |
-
# crop the image around the center based on given height and width
|
41 |
-
def center_crop(img, out_height, out_width):
|
42 |
-
height, width, _ = img.shape
|
43 |
-
left = int((width - out_width) / 2)
|
44 |
-
right = int((width + out_width) / 2)
|
45 |
-
top = int((height - out_height) / 2)
|
46 |
-
bottom = int((height + out_height) / 2)
|
47 |
-
img = img[top:bottom, left:right]
|
48 |
return img
|
49 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
50 |
|
51 |
-
|
|
|
|
|
|
|
|
|
|
|
52 |
|
53 |
-
|
54 |
-
|
55 |
-
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
|
56 |
-
|
57 |
-
img = pre_process_edgetpu(img, (224, 224, 3))
|
58 |
-
|
59 |
-
img_batch = np.expand_dims(img, axis=0)
|
60 |
|
61 |
-
results = sess.run(["Softmax:0"], {"images:0": img_batch})[0]
|
62 |
-
result = reversed(results[0].argsort()[-5:])
|
63 |
-
resultdic = {}
|
64 |
-
for r in result:
|
65 |
-
resultdic[labels[str(r)]] = float(results[0][r])
|
66 |
-
return resultdic
|
67 |
-
|
68 |
-
title="EfficientNet-Lite4"
|
69 |
-
description="EfficientNet-Lite 4 is the largest variant and most accurate of the set of EfficientNet-Lite model. It is an integer-only quantized model that produces the highest accuracy of all of the EfficientNet models. It achieves 80.4% ImageNet top-1 accuracy, while still running in real-time (e.g. 30ms/image) on a Pixel 4 CPU."
|
70 |
examples=[['catonnx.jpg']]
|
71 |
-
gr.Interface(
|
|
|
1 |
+
import mxnet as mx
|
|
|
2 |
import matplotlib.pyplot as plt
|
3 |
+
import numpy as np
|
4 |
+
from collections import namedtuple
|
5 |
+
from mxnet.gluon.data.vision import transforms
|
6 |
+
import os
|
7 |
import gradio as gr
|
|
|
|
|
8 |
|
9 |
+
from PIL import Image
|
10 |
+
import imageio
|
11 |
+
import onnxruntime as ort
|
12 |
|
13 |
+
def get_image(path):
|
14 |
+
'''
|
15 |
+
Using path to image, return the RGB load image
|
16 |
+
'''
|
17 |
+
img = imageio.imread(path, pilmode='RGB')
|
|
|
|
|
|
|
|
|
18 |
return img
|
19 |
|
20 |
+
# Pre-processing function for ImageNet models using numpy
|
21 |
+
def preprocess(img):
|
22 |
+
'''
|
23 |
+
Preprocessing required on the images for inference with mxnet gluon
|
24 |
+
The function takes loaded image and returns processed tensor
|
25 |
+
'''
|
26 |
+
img = np.array(Image.fromarray(img).resize((224, 224))).astype(np.float32)
|
27 |
+
img[:, :, 0] -= 123.68
|
28 |
+
img[:, :, 1] -= 116.779
|
29 |
+
img[:, :, 2] -= 103.939
|
30 |
+
img[:,:,[0,1,2]] = img[:,:,[2,1,0]]
|
31 |
+
img = img.transpose((2, 0, 1))
|
32 |
+
img = np.expand_dims(img, axis=0)
|
33 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
34 |
return img
|
35 |
|
36 |
+
mx.test_utils.download('https://s3.amazonaws.com/model-server/inputs/kitten.jpg')
|
37 |
+
|
38 |
+
mx.test_utils.download('https://s3.amazonaws.com/onnx-model-zoo/synset.txt')
|
39 |
+
with open('synset.txt', 'r') as f:
|
40 |
+
labels = [l.rstrip() for l in f]
|
41 |
+
|
42 |
+
os.system("wget https://github.com/onnx/models/raw/main/vision/classification/inception_and_googlenet/googlenet/model/googlenet-9.onnx")
|
43 |
+
|
44 |
+
ort_session = ort.InferenceSession("googlenet-9.onnx")
|
45 |
+
|
46 |
+
|
47 |
+
def predict(path):
|
48 |
+
img_batch = preprocess(get_image(path))
|
49 |
+
|
50 |
+
outputs = ort_session.run(
|
51 |
+
None,
|
52 |
+
{"data_0": img_batch.astype(np.float32)},
|
53 |
+
)
|
54 |
|
55 |
+
a = np.argsort(-outputs[0].flatten())
|
56 |
+
results = {}
|
57 |
+
for i in a[0:5]:
|
58 |
+
results[labels[i]]=float(outputs[0][0][i])
|
59 |
+
return results
|
60 |
+
|
61 |
|
62 |
+
title="GoogleNet"
|
63 |
+
description="GoogLeNet is the name of a convolutional neural network for classification, which competed in the ImageNet Large Scale Visual Recognition Challenge in 2014."
|
|
|
|
|
|
|
|
|
|
|
64 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
65 |
examples=[['catonnx.jpg']]
|
66 |
+
gr.Interface(predict,gr.inputs.Image(type='filepath'),"label",title=title,description=description,examples=examples).launch(enable_queue=True,debug=True)
|