Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -1,21 +1,11 @@
|
|
1 |
import torch
|
2 |
-
from torch.nn import functional as F
|
3 |
import torchvision.models as models
|
4 |
-
|
|
|
|
|
5 |
import requests
|
6 |
-
|
7 |
-
|
8 |
-
transforms.Resize(256),
|
9 |
-
transforms.CenterCrop(224),
|
10 |
-
transforms.ToTensor(),
|
11 |
-
transforms.Normalize(
|
12 |
-
mean=[0.485, 0.456 , 0.406],
|
13 |
-
std=[0.229, 0.224, 0.225]
|
14 |
-
)
|
15 |
-
])
|
16 |
-
|
17 |
-
response = requests.get("https://git.io/JJkYN")
|
18 |
-
labels = response.text.split("\n")
|
19 |
|
20 |
image_prediction_models = {
|
21 |
'resnet': models.resnet50,
|
@@ -50,26 +40,89 @@ def get_model_names(models_dict):
|
|
50 |
|
51 |
model_list = get_model_names(image_prediction_models)
|
52 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
53 |
def classify_image(input_image, selected_model):
|
54 |
-
|
55 |
-
|
56 |
-
|
57 |
-
|
58 |
-
input_batch = input_batch.to('cuda')
|
59 |
-
|
60 |
-
with torch.no_grad():
|
61 |
-
output = model(input_batch)
|
62 |
-
|
63 |
-
probabilities = F.softmax(input = output[0] , dim = 0)
|
64 |
-
top_prob, top_catid = torch.topk(probabilities, 5)
|
65 |
-
confidences = {labels[top_catid[i].item()]: top_prob[i].item() for i in range(top_prob.size(0))}
|
66 |
-
return confidences
|
67 |
|
68 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
69 |
|
70 |
-
|
71 |
-
fn=classify_image,
|
72 |
-
inputs= [gr.Image(type='pil'),
|
73 |
-
gr.Dropdown(model_list)],
|
74 |
-
outputs=gr.Label(num_top_classes=5))
|
75 |
-
interface.launch()
|
|
|
1 |
import torch
|
|
|
2 |
import torchvision.models as models
|
3 |
+
import torchvision.transforms as transforms
|
4 |
+
import torchvision.datasets as datasets
|
5 |
+
from torchvision.transforms import Compose
|
6 |
import requests
|
7 |
+
import random
|
8 |
+
import gradio as gr
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
9 |
|
10 |
image_prediction_models = {
|
11 |
'resnet': models.resnet50,
|
|
|
40 |
|
41 |
model_list = get_model_names(image_prediction_models)
|
42 |
|
43 |
+
normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
|
44 |
+
|
45 |
+
def preprocess(model_name):
|
46 |
+
input_size = 224
|
47 |
+
if model_name == 'inception':
|
48 |
+
input_size = 299
|
49 |
+
return transforms.Compose([
|
50 |
+
transforms.Resize(256),
|
51 |
+
transforms.CenterCrop(input_size),
|
52 |
+
transforms.ToTensor(),
|
53 |
+
normalize,
|
54 |
+
])
|
55 |
+
|
56 |
+
response = requests.get("https://git.io/JJkYN")
|
57 |
+
labels = response.text.split("\n")
|
58 |
+
|
59 |
+
def postprocess_default(output):
|
60 |
+
probabilities = torch.nn.functional.softmax(output[0], dim=0)
|
61 |
+
top_prob, top_catid = torch.topk(probabilities, 5)
|
62 |
+
confidences = {labels[top_catid[i].item()]: top_prob[i].item() for i in range(top_prob.size(0))}
|
63 |
+
return confidences
|
64 |
+
|
65 |
+
def postprocess_inception(output):
|
66 |
+
probabilities = torch.nn.functional.softmax(output[1], dim=0)
|
67 |
+
top_prob, top_catid = torch.topk(probabilities, 5)
|
68 |
+
confidences = {labels[top_catid[i].item()]: top_prob[i].item() for i in range(top_prob.size(0))}
|
69 |
+
return confidences
|
70 |
+
|
71 |
def classify_image(input_image, selected_model):
|
72 |
+
preprocess_input = preprocess(model_name=selected_model)
|
73 |
+
input_tensor = preprocess_input(input_image)
|
74 |
+
input_batch = input_tensor.unsqueeze(0)
|
75 |
+
model = load_pretrained_model(selected_model)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
76 |
|
77 |
+
if torch.cuda.is_available():
|
78 |
+
input_batch = input_batch.to('cuda')
|
79 |
+
model.to('cuda')
|
80 |
+
|
81 |
+
model.eval()
|
82 |
+
with torch.no_grad():
|
83 |
+
output = model(input_batch)
|
84 |
+
|
85 |
+
if selected_model.lower() == 'inception':
|
86 |
+
return postprocess_inception(output)
|
87 |
+
else:
|
88 |
+
return postprocess_default(output)
|
89 |
+
|
90 |
+
def get_random_image():
|
91 |
+
cifar10 = datasets.CIFAR10(root='./data', train=False, download=True, transform=transforms.ToTensor())
|
92 |
+
random_idx = random.randint(0, len(cifar10) - 1)
|
93 |
+
image, _ = cifar10[random_idx]
|
94 |
+
image = transforms.ToPILImage()(image)
|
95 |
+
return image
|
96 |
+
|
97 |
+
def generate_random_image():
|
98 |
+
image = get_random_image()
|
99 |
+
return image
|
100 |
+
|
101 |
+
def classify_generated_image(image, model):
|
102 |
+
return classify_image(image, model)
|
103 |
+
|
104 |
+
with gr.Blocks() as demo:
|
105 |
+
with gr.Tabs():
|
106 |
+
with gr.TabItem("Upload Image"):
|
107 |
+
with gr.Row():
|
108 |
+
with gr.Column():
|
109 |
+
upload_image = gr.Image(type='pil', label="Upload Image")
|
110 |
+
model_dropdown_upload = gr.Dropdown(model_list, label="Select Model")
|
111 |
+
classify_button_upload = gr.Button("Classify")
|
112 |
+
with gr.Column():
|
113 |
+
output_label_upload = gr.Label(num_top_classes=5)
|
114 |
+
classify_button_upload.click(classify_image, inputs=[upload_image, model_dropdown_upload], outputs=output_label_upload)
|
115 |
+
|
116 |
+
with gr.TabItem("Generate Random Image"):
|
117 |
+
with gr.Row():
|
118 |
+
with gr.Column():
|
119 |
+
generate_button = gr.Button("Generate Random Image")
|
120 |
+
random_image_output = gr.Image(type='pil', label="Random CIFAR-10 Image")
|
121 |
+
with gr.Column():
|
122 |
+
model_dropdown_random = gr.Dropdown(model_list, label="Select Model")
|
123 |
+
classify_button_random = gr.Button("Classify")
|
124 |
+
output_label_random = gr.Label(num_top_classes=5)
|
125 |
+
generate_button.click(generate_random_image, inputs=[], outputs=random_image_output)
|
126 |
+
classify_button_random.click(classify_generated_image, inputs=[random_image_output, model_dropdown_random], outputs=output_label_random)
|
127 |
|
128 |
+
demo.launch()
|
|
|
|
|
|
|
|
|
|