Spaces:
Sleeping
Sleeping
File size: 13,580 Bytes
77da9e2 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 |
"""
Shared Gradio Interface Factory
This module provides a reusable Gradio interface factory that works with
different detection backends (direct service or API client).
This eliminates code duplication between app.py and ui/gradio_interface.py
"""
import gradio as gr
from typing import Callable, Optional
def _handle_ocr_only_toggle(is_ocr_only: bool):
"""
Update dependent controls when OCR-only mode is toggled.
Returns tuple of updates for:
- CLIP checkbox
- OCR checkbox
- BLIP checkbox
- BLIP scope radio
"""
if is_ocr_only:
return (
gr.update(value=False, interactive=False),
gr.update(value=True, interactive=False),
gr.update(value=False, interactive=False),
gr.update(value="Only image & button", visible=False),
)
return (
gr.update(interactive=True),
gr.update(value=True, interactive=True),
gr.update(interactive=True),
gr.update(visible=False),
)
def create_interface(
detection_fn: Callable,
title_suffix: str = "",
show_api_info: bool = False,
api_url: Optional[str] = None
) -> gr.Blocks:
"""
Create a Gradio interface with a pluggable detection function
Args:
detection_fn: Function that takes (image, confidence, thickness, clip, ocr, blip, ocr_only, blip_scope)
and returns (annotated_image, summary, json_data)
title_suffix: Additional text for the title
show_api_info: Whether to show API connection info
api_url: API URL to display (if show_api_info=True)
Returns:
Gradio Blocks interface
"""
with gr.Blocks(title="CU-1 UI Element Detector", theme=gr.themes.Soft()) as interface:
# Build title markdown
title_parts = [
"# 🎯 CU-1 UI Element Detector",
"",
"Detect interactive elements in screenshots and UI mockups.",
"",
"**Multi-Model Pipeline:**",
"- 🔍 **RF-DETR** detects all UI elements (single class detection)",
"- 🏷️ **CLIP** classifies elements into 6 types (button, input, text, image, list_item, navigation)",
"- 📝 **OCR** extracts text content from detected elements",
"- 🖼️ **BLIP** generates visual descriptions for icons"
]
if title_suffix:
title_parts.append("")
title_parts.append(f"**{title_suffix}**")
if show_api_info and api_url:
title_parts.append("")
title_parts.append(f"**API:** Connected to `{api_url}`")
gr.Markdown("\n".join(title_parts))
with gr.Row():
with gr.Column(scale=1):
input_image = gr.Image(
type="pil",
label="Upload Screenshot",
height=400,
sources=["upload"]
)
with gr.Accordion("Detection Settings", open=True):
confidence_slider = gr.Slider(
minimum=0.1,
maximum=0.9,
value=0.35,
step=0.05,
label="Confidence Threshold",
info="Lower = more elements detected"
)
thickness_slider = gr.Slider(
minimum=1,
maximum=6,
value=2,
step=1,
label="Box Line Thickness"
)
with gr.Accordion("Feature Settings", open=True):
clip_checkbox = gr.Checkbox(
value=False,
label="Enable CLIP Classification",
info="Classify elements into types (slower but more informative)"
)
ocr_checkbox = gr.Checkbox(
value=True,
label="Enable OCR Text Extraction",
info="Extract text content from elements"
)
blip_checkbox = gr.Checkbox(
value=False,
label="Enable BLIP Description",
info="Generate visual descriptions for icons (slower)"
)
ocr_only_checkbox = gr.Checkbox(
value=False,
label="OCR-only (skip detection/classification)",
info="Run OCR across the whole image and return OCR boxes only"
)
blip_scope_radio = gr.Radio(
choices=["Only image & button", "All elements"],
value="Only image & button",
label="BLIP Scope",
info="When to apply BLIP descriptions",
visible=False
)
with gr.Accordion("🎨 Preprocessing (Cross-Device Consistency)", open=False):
preprocess_checkbox = gr.Checkbox(
value=False,
label="Enable Image Preprocessing",
info="Standardize screenshots from different devices (Samsung, Pixel, Oppo, etc.)"
)
preprocess_mode_radio = gr.Radio(
choices=["RF-DETR Optimized (Recommended)", "Generic (CLIP/OCR Focus)"],
value="RF-DETR Optimized (Recommended)",
label="Preprocessing Mode",
info="RF-DETR: Preserves ImageNet normalization | Generic: Aggressive for OCR",
visible=False
)
preprocess_preset_dropdown = gr.Dropdown(
choices=["gentle", "standard", "aggressive_denoise", "color_only"],
value="standard",
label="Preprocessing Preset",
info="gentle=minimal | standard=balanced | aggressive_denoise=strong | color_only=colors",
visible=False
)
detect_button = gr.Button("🔍 Detect Elements", variant="primary", size="lg")
with gr.Column(scale=1):
output_image = gr.Image(
type="pil",
label="Detected Elements",
height=400
)
summary_output = gr.Markdown(label="Detection Summary")
with gr.Accordion("Raw Results (JSON)", open=False):
json_output = gr.JSON(label="Detections JSON")
with gr.Accordion("API Quickstart", open=False):
api_docs = gr.Markdown(
value="\n".join([
"#### Call the Detection API",
"",
"```bash",
"curl -X POST \"https://your-space.hf.space/detect\" \\",
" -H \"Authorization: Bearer <HF_TOKEN>\" \\",
" -F \"image=@screenshot.png\" \\",
" -F \"confidence_threshold=0.35\" \\",
" -F \"enable_clip=true\" \\",
" -F \"enable_ocr=true\"",
"```",
"",
"```python",
"import requests",
"",
"url = \"https://your-space.hf.space/detect\"",
"headers = {\"Authorization\": \"Bearer <HF_TOKEN>\"}",
"files = {\"image\": open(\"screenshot.png\", \"rb\")}",
"data = {",
" \"confidence_threshold\": 0.35,",
" \"enable_clip\": \"true\",",
" \"enable_ocr\": \"true\"",
"}",
"resp = requests.post(url, files=files, data=data, headers=headers, timeout=120)",
"resp.raise_for_status()",
"print(resp.json())",
"```",
"",
"- Replace `your-space` with your Hugging Face Space slug.",
"- Add the `Authorization` header for private Spaces.",
"- Response payload includes bounding boxes, texts, and optional annotated image."
])
)
# Toggle BLIP scope visibility
blip_checkbox.change(
fn=lambda v: gr.update(visible=v),
inputs=blip_checkbox,
outputs=blip_scope_radio
)
# Handle OCR-only toggle to disable/enable related controls
ocr_only_checkbox.change(
fn=_handle_ocr_only_toggle,
inputs=ocr_only_checkbox,
outputs=[clip_checkbox, ocr_checkbox, blip_checkbox, blip_scope_radio]
)
# Toggle preprocessing options visibility
def toggle_preprocess_options(enabled):
return gr.update(visible=enabled), gr.update(visible=enabled)
preprocess_checkbox.change(
fn=toggle_preprocess_options,
inputs=preprocess_checkbox,
outputs=[preprocess_mode_radio, preprocess_preset_dropdown]
)
# Update preset choices based on mode
def update_preset_choices(mode):
if "RF-DETR" in mode:
return gr.update(
choices=["gentle", "standard", "aggressive_denoise", "color_only"],
value="standard",
info="gentle=minimal | standard=balanced | aggressive_denoise=strong | color_only=colors"
)
else: # Generic mode
return gr.update(
choices=["minimal", "standard", "aggressive", "ocr_optimized"],
value="standard",
info="minimal=light | standard=balanced | aggressive=maximum | ocr_optimized=best for text"
)
preprocess_mode_radio.change(
fn=update_preset_choices,
inputs=preprocess_mode_radio,
outputs=preprocess_preset_dropdown
)
# Connect detection button
detect_button.click(
fn=detection_fn,
inputs=[
input_image,
confidence_slider,
thickness_slider,
clip_checkbox,
ocr_checkbox,
blip_checkbox,
ocr_only_checkbox,
blip_scope_radio,
preprocess_checkbox,
preprocess_mode_radio,
preprocess_preset_dropdown
],
outputs=[output_image, summary_output, json_output]
)
# Build footer markdown
footer_parts = [
"---",
"### ⚡ Performance Tips",
"",
"- **Fast mode** (CLIP ❌, OCR ✅): ~30-40s - Good for text extraction",
"- **Balanced mode** (CLIP ✅, OCR ✅): ~50-60s - Full classification + text",
"- **Ultra-fast mode** (CLIP ❌, OCR ❌): ~25-35s - Just bounding boxes",
"",
"### 🎨 Cross-Device Preprocessing",
"",
"Testing on multiple devices (Samsung, Pixel, Oppo)? **Enable preprocessing** for consistent results!",
"",
"- **RF-DETR Optimized** (Recommended): Preserves ImageNet normalization, best for detection",
"- **Generic Mode**: Aggressive normalization, best for OCR accuracy",
"",
"### 🏗️ Architecture",
"",
"**Single-Class Detection:** RF-DETR detects generic \"UI elements\" (one class)",
"**Multi-Class Classification:** CLIP classifies detections into 6 specific types"
]
if show_api_info and api_url:
footer_parts.extend([
"",
"### 🔧 API Connection",
"",
f"This UI is a **client** of the API server at `{api_url}`",
"",
"**Communication:** HTTP/REST (multipart/form-data)",
"**Separation:** UI layer is completely isolated from detection logic",
"",
"To change API endpoint:",
"```bash",
"export CU1_API_URL=http://your-api-server:8000",
"python app_ui.py",
"```"
])
else:
footer_parts.extend([
"",
"### 📦 Deployment",
"",
"This app uses direct detection service access (no API layer).",
"Optimized for Hugging Face Spaces and local testing."
])
gr.Markdown("\n".join(footer_parts))
return interface
|