Spaces:
Sleeping
Sleeping
pradanaadn
commited on
Commit
·
869b2b3
1
Parent(s):
14a6bf2
feat: create ui using gradio
Browse files- .gitignore +3 -0
- .python-version +1 -0
- examples/cardbox.jpeg +0 -0
- examples/glass.jpeg +0 -0
- examples/plastic.png +0 -0
- main.py +109 -0
- pyproject.toml +13 -0
- requirements.txt +12 -0
.gitignore
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
.venv
|
2 |
+
.gradio
|
3 |
+
uv.lock
|
.python-version
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
3.12
|
examples/cardbox.jpeg
ADDED
examples/glass.jpeg
ADDED
examples/plastic.png
ADDED
main.py
ADDED
@@ -0,0 +1,109 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
import torch
|
3 |
+
import torch.nn as nn
|
4 |
+
from huggingface_hub import PyTorchModelHubMixin
|
5 |
+
from torchvision.models import mobilenet_v3_large
|
6 |
+
from torchvision.transforms import v2
|
7 |
+
from PIL import Image
|
8 |
+
import os
|
9 |
+
|
10 |
+
|
11 |
+
class TrashMobileNet(nn.Module, PyTorchModelHubMixin):
|
12 |
+
def __init__(self, num_classes=6):
|
13 |
+
super(TrashMobileNet, self).__init__()
|
14 |
+
self.model = mobilenet_v3_large(weights="DEFAULT")
|
15 |
+
for param in self.model.parameters():
|
16 |
+
param.requires_grad = False
|
17 |
+
num_features = self.model.classifier[-1].in_features
|
18 |
+
self.model.classifier[-1] = nn.Linear(num_features, num_classes)
|
19 |
+
for param in self.model.classifier[-1].parameters():
|
20 |
+
param.requires_grad = True
|
21 |
+
|
22 |
+
def forward(self, x):
|
23 |
+
x = self.model(x)
|
24 |
+
return x
|
25 |
+
|
26 |
+
|
27 |
+
# Load the model from Hugging Face Hub
|
28 |
+
model_name = "pradanaadn/trash-clasification"
|
29 |
+
model = TrashMobileNet.from_pretrained(model_name)
|
30 |
+
model.eval()
|
31 |
+
|
32 |
+
# Define the image transformations
|
33 |
+
transform = v2.Compose([
|
34 |
+
v2.Resize((224, 224)),
|
35 |
+
v2.ToImage(),
|
36 |
+
v2.ToDtype(torch.float32, scale=True),
|
37 |
+
])
|
38 |
+
|
39 |
+
|
40 |
+
def predict(image):
|
41 |
+
"""
|
42 |
+
Prediction function that takes a Gradio image input and returns class probabilities
|
43 |
+
"""
|
44 |
+
labels = ["cardboard", "glass", "metal", "paper", "plastic", "trash"]
|
45 |
+
|
46 |
+
# Convert Gradio image to PIL Image if it's not already
|
47 |
+
if not isinstance(image, Image.Image):
|
48 |
+
image = Image.fromarray(image)
|
49 |
+
|
50 |
+
# Apply transformations and add batch dimension
|
51 |
+
image_tensor = transform(image)
|
52 |
+
image_tensor = image_tensor.unsqueeze(0)
|
53 |
+
|
54 |
+
# Get model predictions
|
55 |
+
with torch.no_grad():
|
56 |
+
outputs = model(image_tensor)
|
57 |
+
probabilities = torch.nn.functional.softmax(outputs, dim=1)
|
58 |
+
probabilities = probabilities[0].tolist()
|
59 |
+
|
60 |
+
# Create dictionary of label-probability pairs
|
61 |
+
return {label: float(prob) for label, prob in zip(labels, probabilities)}
|
62 |
+
|
63 |
+
|
64 |
+
|
65 |
+
|
66 |
+
# Create example images if they don't exist (you would need to provide these images)
|
67 |
+
examples = [
|
68 |
+
["examples/cardbox.jpeg", "A cardboard box"],
|
69 |
+
["examples/glass.jpeg", "A glass bottle"],
|
70 |
+
["examples/plastic.png", "Mixed trash"]
|
71 |
+
]
|
72 |
+
|
73 |
+
|
74 |
+
with gr.Blocks() as iface:
|
75 |
+
|
76 |
+
|
77 |
+
with gr.Row():
|
78 |
+
with gr.Column():
|
79 |
+
input_image = gr.Image(
|
80 |
+
label="Upload Image",
|
81 |
+
type="pil",
|
82 |
+
elem_id="image_upload"
|
83 |
+
)
|
84 |
+
submit_btn = gr.Button("Classify", variant="primary")
|
85 |
+
|
86 |
+
with gr.Column():
|
87 |
+
output_label = gr.Label(
|
88 |
+
label="Classification Results",
|
89 |
+
num_top_classes=6
|
90 |
+
)
|
91 |
+
|
92 |
+
gr.Markdown("### Example Images")
|
93 |
+
gr.Examples(
|
94 |
+
examples=examples,
|
95 |
+
inputs=input_image,
|
96 |
+
outputs=output_label,
|
97 |
+
fn=predict,
|
98 |
+
cache_examples=True
|
99 |
+
)
|
100 |
+
|
101 |
+
submit_btn.click(
|
102 |
+
fn=predict,
|
103 |
+
inputs=input_image,
|
104 |
+
outputs=output_label
|
105 |
+
)
|
106 |
+
|
107 |
+
|
108 |
+
# Launch the interface
|
109 |
+
iface.launch(share=True)
|
pyproject.toml
ADDED
@@ -0,0 +1,13 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
[project]
|
2 |
+
name = "trash-classification"
|
3 |
+
version = "0.1.0"
|
4 |
+
description = "Add your description here"
|
5 |
+
readme = "README.md"
|
6 |
+
requires-python = ">=3.12"
|
7 |
+
dependencies = [
|
8 |
+
"gradio>=5.3.0",
|
9 |
+
"huggingface-hub>=0.27.0",
|
10 |
+
"torch>=2.5.1",
|
11 |
+
"torchvision>=0.20.1",
|
12 |
+
"transformers>=4.47.1",
|
13 |
+
]
|
requirements.txt
ADDED
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# This file was autogenerated by uv via the following command:
|
2 |
+
# uv pip compile pyproject.toml --no-deps -o requirements.txt
|
3 |
+
gradio==5.9.1
|
4 |
+
# via trash-classification (pyproject.toml)
|
5 |
+
huggingface-hub==0.27.0
|
6 |
+
# via trash-classification (pyproject.toml)
|
7 |
+
torch==2.5.1
|
8 |
+
# via trash-classification (pyproject.toml)
|
9 |
+
torchvision==0.20.1
|
10 |
+
# via trash-classification (pyproject.toml)
|
11 |
+
transformers==4.47.1
|
12 |
+
# via trash-classification (pyproject.toml)
|