Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,111 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from PIL import Image, ImageDraw, ImageFont
|
2 |
+
import gradio as gr
|
3 |
+
from smolagents import CodeAgent, InferenceClientModel, DuckDuckGoSearchTool, Tool
|
4 |
+
from gradio_client import Client
|
5 |
+
|
6 |
+
#%% Tool Wrapper for the Hugging Face Space
|
7 |
+
class TextToImageTool(Tool):
|
8 |
+
name = "text_to_image"
|
9 |
+
description = "Generate an image from a text prompt using m-ric/text-to-image."
|
10 |
+
|
11 |
+
def __init__(self):
|
12 |
+
super().__init__()
|
13 |
+
self.client = Client("m-ric/text-to-image") # Calls HF Space
|
14 |
+
|
15 |
+
def run(self, prompt: str):
|
16 |
+
image = self.client.predict(prompt, api_name="/predict")
|
17 |
+
return image # This is a PIL image
|
18 |
+
|
19 |
+
#%% Utility functions
|
20 |
+
def add_label_to_image(image, label):
|
21 |
+
draw = ImageDraw.Draw(image)
|
22 |
+
font_path = "/usr/share/fonts/truetype/dejavu/DejaVuSans-Bold.ttf"
|
23 |
+
font_size = 30
|
24 |
+
try:
|
25 |
+
font = ImageFont.truetype(font_path, font_size)
|
26 |
+
except:
|
27 |
+
font = ImageFont.load_default()
|
28 |
+
text_bbox = draw.textbbox((0, 0), label, font=font)
|
29 |
+
text_width = text_bbox[2] - text_bbox[0]
|
30 |
+
text_height = text_bbox[3] - text_bbox[1]
|
31 |
+
position = (image.width - text_width - 20, image.height - text_height - 20)
|
32 |
+
rect_margin = 10
|
33 |
+
rect_position = [
|
34 |
+
position[0] - rect_margin, position[1] - rect_margin,
|
35 |
+
position[0] + text_width + rect_margin, position[1] + text_height + rect_margin
|
36 |
+
]
|
37 |
+
draw.rectangle(rect_position, fill=(0, 0, 0, 128))
|
38 |
+
draw.text(position, label, fill="white", font=font)
|
39 |
+
return image
|
40 |
+
|
41 |
+
def plot_and_save_agent_image(image, label, save_path=None):
|
42 |
+
labeled_image = add_label_to_image(image, label)
|
43 |
+
labeled_image.show()
|
44 |
+
if save_path:
|
45 |
+
labeled_image.save(save_path)
|
46 |
+
print(f"Image saved to {save_path}")
|
47 |
+
|
48 |
+
def generate_prompts_for_object(object_name):
|
49 |
+
return {
|
50 |
+
"past": f"Show an old version of a {object_name} from its early days.",
|
51 |
+
"present": f"Show a {object_name} with current features/design/technology.",
|
52 |
+
"future": f"Show a futuristic version of a {object_name}, by predicting advanced features and futuristic design."
|
53 |
+
}
|
54 |
+
|
55 |
+
def generate_object_history(object_name):
|
56 |
+
prompts = generate_prompts_for_object(object_name)
|
57 |
+
labels = {
|
58 |
+
"past": f"{object_name} - Past",
|
59 |
+
"present": f"{object_name} - Present",
|
60 |
+
"future": f"{object_name} - Future"
|
61 |
+
}
|
62 |
+
|
63 |
+
images = []
|
64 |
+
for time_period, prompt in prompts.items():
|
65 |
+
print(f"Generating {time_period} frame: {prompt}")
|
66 |
+
result = agent.run(prompt) # Runs tool
|
67 |
+
if hasattr(result, "to_raw"): # If wrapped output
|
68 |
+
result = result.to_raw()
|
69 |
+
images.append(result)
|
70 |
+
plot_and_save_agent_image(result, labels[time_period], save_path=f"{object_name}_{time_period}.png")
|
71 |
+
|
72 |
+
gif_path = f"{object_name}_evolution.gif"
|
73 |
+
images[0].save(gif_path, save_all=True, append_images=images[1:], duration=1000, loop=0)
|
74 |
+
return images, gif_path
|
75 |
+
|
76 |
+
#%% Tool & Agent Setup
|
77 |
+
image_generation_tool = TextToImageTool()
|
78 |
+
search_tool = DuckDuckGoSearchTool()
|
79 |
+
llm_engine = InferenceClientModel("Qwen/Qwen2.5-72B-Instruct")
|
80 |
+
|
81 |
+
agent = CodeAgent(tools=[image_generation_tool, search_tool], model=llm_engine)
|
82 |
+
|
83 |
+
#%% Gradio Interface
|
84 |
+
def create_gradio_interface():
|
85 |
+
with gr.Blocks() as demo:
|
86 |
+
gr.Markdown("# TimeMetamorphy: an object Evolution Generator")
|
87 |
+
gr.Markdown("""
|
88 |
+
## Unlocking the secrets of time!
|
89 |
+
Enter an object name (like bicycle or smartphone), and this app will generate its visual evolution.
|
90 |
+
""")
|
91 |
+
|
92 |
+
default_images = [
|
93 |
+
("car_past.png", "Car - Past"),
|
94 |
+
("car_present.png", "Car - Present"),
|
95 |
+
("car_future.png", "Car - Future")
|
96 |
+
]
|
97 |
+
default_gif_path = "car_evolution.gif"
|
98 |
+
|
99 |
+
with gr.Row():
|
100 |
+
with gr.Column():
|
101 |
+
object_name_input = gr.Textbox(label="Enter an object name", placeholder="e.g., bicycle, phone")
|
102 |
+
generate_button = gr.Button("Generate Evolution")
|
103 |
+
image_gallery = gr.Gallery(label="Generated Images", columns=3, rows=1, value=default_images)
|
104 |
+
gif_output = gr.Image(label="Generated GIF", value=default_gif_path)
|
105 |
+
|
106 |
+
generate_button.click(fn=generate_object_history, inputs=[object_name_input], outputs=[image_gallery, gif_output])
|
107 |
+
return demo
|
108 |
+
|
109 |
+
# Launch app
|
110 |
+
demo = create_gradio_interface()
|
111 |
+
demo.launch(share=True)
|