Spaces:
Sleeping
Sleeping
yahiab
commited on
Commit
·
0d11696
1
Parent(s):
882271f
Add public model testing app
Browse files- app.py +105 -0
- requirements.txt +4 -0
app.py
ADDED
@@ -0,0 +1,105 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
import numpy as np
|
3 |
+
from PIL import Image, ImageDraw
|
4 |
+
import torch
|
5 |
+
from torchvision import transforms
|
6 |
+
from transformers import AutoModelForImageClassification, AutoFeatureExtractor
|
7 |
+
|
8 |
+
# Define all available models
|
9 |
+
MODEL_LIST = {
|
10 |
+
'beit': "microsoft/beit-base-patch16-224-pt22k-ft22k",
|
11 |
+
'vit': "google/vit-base-patch16-224",
|
12 |
+
'convnext': "facebook/convnext-tiny-224",
|
13 |
+
}
|
14 |
+
|
15 |
+
# Preprocessing transforms
|
16 |
+
def get_preprocessor(model_name):
|
17 |
+
extractor = AutoFeatureExtractor.from_pretrained(MODEL_LIST[model_name])
|
18 |
+
return extractor
|
19 |
+
|
20 |
+
# Load a model from Hugging Face
|
21 |
+
def load_model(model_name):
|
22 |
+
model = AutoModelForImageClassification.from_pretrained(MODEL_LIST[model_name]).cuda().eval()
|
23 |
+
return model
|
24 |
+
|
25 |
+
# Function to make predictions
|
26 |
+
def predict(image, model, preprocessor):
|
27 |
+
inputs = preprocessor(images=image, return_tensors="pt").to("cuda")
|
28 |
+
with torch.no_grad():
|
29 |
+
outputs = model(**inputs)
|
30 |
+
predicted_class = torch.argmax(outputs.logits, dim=1).item()
|
31 |
+
return model.config.id2label[predicted_class]
|
32 |
+
|
33 |
+
# Function to draw a rectangle on the image
|
34 |
+
def draw_rectangle(image, x, y, size=224):
|
35 |
+
image_pil = image.copy() # Create a copy to avoid modifying the original image
|
36 |
+
draw = ImageDraw.Draw(image_pil)
|
37 |
+
x1, y1 = x, y
|
38 |
+
x2, y2 = x + size, y + size
|
39 |
+
draw.rectangle([x1, y1, x2, y2], outline="red", width=5)
|
40 |
+
return image_pil
|
41 |
+
|
42 |
+
# Function to crop the image
|
43 |
+
def crop_image(image, x, y, size=224):
|
44 |
+
image_np = np.array(image)
|
45 |
+
h, w, _ = image_np.shape
|
46 |
+
x = min(max(x, 0), w - size)
|
47 |
+
y = min(max(y, 0), h - size)
|
48 |
+
cropped = image_np[y:y+size, x:x+size]
|
49 |
+
return Image.fromarray(cropped)
|
50 |
+
|
51 |
+
# Global variables
|
52 |
+
current_model = None
|
53 |
+
current_preprocessor = None
|
54 |
+
|
55 |
+
# Gradio Interface
|
56 |
+
with gr.Blocks() as demo:
|
57 |
+
gr.Markdown("## Test Public Models for Coral Classification")
|
58 |
+
|
59 |
+
with gr.Row():
|
60 |
+
with gr.Column():
|
61 |
+
model_selector = gr.Dropdown(choices=list(MODEL_LIST.keys()), value='beit', label="Select Model")
|
62 |
+
image_input = gr.Image(type="pil", label="Upload Image", interactive=True)
|
63 |
+
x_slider = gr.Slider(minimum=0, maximum=1000, step=1, value=0, label="X Coordinate")
|
64 |
+
y_slider = gr.Slider(minimum=0, maximum=1000, step=1, value=0, label="Y Coordinate")
|
65 |
+
with gr.Column():
|
66 |
+
interactive_image = gr.Image(label="Interactive Image with Selection")
|
67 |
+
cropped_image = gr.Image(label="Cropped Patch")
|
68 |
+
label_output = gr.Textbox(label="Predicted Label")
|
69 |
+
|
70 |
+
# Update the current model and preprocessor
|
71 |
+
def update_model(model_name):
|
72 |
+
global current_model, current_preprocessor
|
73 |
+
current_model = load_model(model_name)
|
74 |
+
current_preprocessor = get_preprocessor(model_name)
|
75 |
+
return f"Model {model_name} loaded successfully."
|
76 |
+
|
77 |
+
# Update the rectangle and crop the patch
|
78 |
+
def update_selection(image, x, y):
|
79 |
+
overlay_image = draw_rectangle(image, x, y)
|
80 |
+
cropped = crop_image(image, x, y)
|
81 |
+
return overlay_image, cropped
|
82 |
+
|
83 |
+
# Predict the label from the cropped patch
|
84 |
+
def predict_from_cropped(cropped):
|
85 |
+
return predict(cropped, current_model, current_preprocessor)
|
86 |
+
|
87 |
+
# Buttons and interactions
|
88 |
+
crop_button = gr.Button("Crop")
|
89 |
+
crop_button.click(fn=update_selection, inputs=[image_input, x_slider, y_slider], outputs=[interactive_image, cropped_image])
|
90 |
+
|
91 |
+
predict_button = gr.Button("Predict")
|
92 |
+
predict_button.click(fn=predict_from_cropped, inputs=cropped_image, outputs=label_output)
|
93 |
+
|
94 |
+
model_selector.change(fn=update_model, inputs=model_selector, outputs=None)
|
95 |
+
|
96 |
+
# Update sliders dynamically based on uploaded image size
|
97 |
+
def update_sliders(image):
|
98 |
+
if image is not None:
|
99 |
+
width, height = image.size
|
100 |
+
return gr.update(maximum=width - 224), gr.update(maximum=height - 224)
|
101 |
+
return gr.update(), gr.update()
|
102 |
+
|
103 |
+
image_input.change(fn=update_sliders, inputs=image_input, outputs=[x_slider, y_slider])
|
104 |
+
|
105 |
+
demo.launch(server_name="0.0.0.0", server_port=7860)
|
requirements.txt
ADDED
@@ -0,0 +1,4 @@
|
|
|
|
|
|
|
|
|
|
|
1 |
+
gradio
|
2 |
+
torch
|
3 |
+
torchvision
|
4 |
+
transformers
|