Spaces:
Sleeping
Sleeping
mochuan zhan commited on
Commit ·
dfefec8
1
Parent(s): fceab91
fix again
Browse files
app.py
CHANGED
|
@@ -1,8 +1,9 @@
|
|
| 1 |
import gradio as gr
|
| 2 |
import torch
|
| 3 |
import torchvision.transforms as transforms
|
| 4 |
-
from PIL import Image
|
| 5 |
import torch.nn as nn
|
|
|
|
| 6 |
|
| 7 |
# 如果你的模型结构与标准的torchvision模型不同,请确保在此处定义或导入你的模型结构
|
| 8 |
# 例如,如果你有一个model.py文件:
|
|
@@ -79,19 +80,31 @@ transform = transforms.Compose([
|
|
| 79 |
|
| 80 |
# 定义预测函数
|
| 81 |
def classify_image(image):
|
| 82 |
-
#
|
| 83 |
-
|
| 84 |
-
image = image.convert("RGB")
|
| 85 |
|
| 86 |
-
#
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 87 |
img = transform(image).unsqueeze(0) # 添加批次维度
|
| 88 |
|
| 89 |
# 模型预测
|
| 90 |
with torch.no_grad():
|
| 91 |
outputs = model(img)
|
|
|
|
|
|
|
| 92 |
|
| 93 |
# 获取预测结果
|
| 94 |
_, predicted = torch.max(outputs, 1)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 95 |
return str(predicted.item())
|
| 96 |
|
| 97 |
# # 创建Gradio界面
|
|
@@ -103,16 +116,13 @@ def classify_image(image):
|
|
| 103 |
# description="上传一张28x28的灰度图像,模型将预测其所属的数字类别。"
|
| 104 |
# )
|
| 105 |
|
| 106 |
-
|
| 107 |
iface = gr.Interface(
|
| 108 |
fn=classify_image,
|
| 109 |
-
inputs=gr.Sketchpad(
|
| 110 |
-
shape=(224, 224),
|
| 111 |
-
label="Draw a digit"
|
| 112 |
-
),
|
| 113 |
outputs=gr.Label(num_top_classes=1),
|
| 114 |
title="MNIST Digit Classification with ViT",
|
| 115 |
description="使用鼠标手绘一个数字,模型将预测其所属的类别。"
|
| 116 |
)
|
| 117 |
|
|
|
|
| 118 |
iface.launch()
|
|
|
|
| 1 |
import gradio as gr
|
| 2 |
import torch
|
| 3 |
import torchvision.transforms as transforms
|
| 4 |
+
from PIL import Image, ImageOps
|
| 5 |
import torch.nn as nn
|
| 6 |
+
import torch.nn.functional as F
|
| 7 |
|
| 8 |
# 如果你的模型结构与标准的torchvision模型不同,请确保在此处定义或导入你的模型结构
|
| 9 |
# 例如,如果你有一个model.py文件:
|
|
|
|
| 80 |
|
| 81 |
# 定义预测函数
|
| 82 |
def classify_image(image):
|
| 83 |
+
# 将 NumPy 数组转换为 PIL 图像
|
| 84 |
+
image = Image.fromarray(image).convert("L")
|
|
|
|
| 85 |
|
| 86 |
+
# 反转颜色
|
| 87 |
+
image = ImageOps.invert(image)
|
| 88 |
+
|
| 89 |
+
# 调整图像大小到模型需要的输入尺寸
|
| 90 |
+
image = image.resize((224, 224))
|
| 91 |
+
|
| 92 |
+
# 图像预处理(根据您的模型需要进行调整)
|
| 93 |
img = transform(image).unsqueeze(0) # 添加批次维度
|
| 94 |
|
| 95 |
# 模型预测
|
| 96 |
with torch.no_grad():
|
| 97 |
outputs = model(img)
|
| 98 |
+
# 如果模型输出未经过 softmax,可以添加
|
| 99 |
+
probabilities = F.softmax(outputs, dim=1)
|
| 100 |
|
| 101 |
# 获取预测结果
|
| 102 |
_, predicted = torch.max(outputs, 1)
|
| 103 |
+
|
| 104 |
+
# 如果需要返回概率
|
| 105 |
+
# return {str(predicted.item()): probabilities[0][predicted].item()}
|
| 106 |
+
|
| 107 |
+
# 只返回预测的类别
|
| 108 |
return str(predicted.item())
|
| 109 |
|
| 110 |
# # 创建Gradio界面
|
|
|
|
| 116 |
# description="上传一张28x28的灰度图像,模型将预测其所属的数字类别。"
|
| 117 |
# )
|
| 118 |
|
|
|
|
| 119 |
iface = gr.Interface(
|
| 120 |
fn=classify_image,
|
| 121 |
+
inputs=gr.Sketchpad(crop_size=(256,256), type='numpy', image_mode='L', brush=gr.Brush()),
|
|
|
|
|
|
|
|
|
|
| 122 |
outputs=gr.Label(num_top_classes=1),
|
| 123 |
title="MNIST Digit Classification with ViT",
|
| 124 |
description="使用鼠标手绘一个数字,模型将预测其所属的类别。"
|
| 125 |
)
|
| 126 |
|
| 127 |
+
|
| 128 |
iface.launch()
|