Spaces:
Running
Running
Synced repo using 'sync_with_huggingface' Github Action
Browse files- gradio_app.py +254 -0
- requirements.txt +4 -0
gradio_app.py
ADDED
|
@@ -0,0 +1,254 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import os
|
| 2 |
+
import sys
|
| 3 |
+
|
| 4 |
+
if "APP_PATH" in os.environ:
|
| 5 |
+
os.chdir(os.environ["APP_PATH"])
|
| 6 |
+
# fix sys.path for import
|
| 7 |
+
sys.path.append(os.getcwd())
|
| 8 |
+
|
| 9 |
+
from typing import List
|
| 10 |
+
|
| 11 |
+
import gradio as gr
|
| 12 |
+
import pypdfium2
|
| 13 |
+
from pypdfium2 import PdfiumError
|
| 14 |
+
|
| 15 |
+
from surya.detection import batch_text_detection
|
| 16 |
+
from surya.input.pdflines import get_page_text_lines, get_table_blocks
|
| 17 |
+
from surya.layout import batch_layout_detection
|
| 18 |
+
from surya.model.detection.model import load_model, load_processor
|
| 19 |
+
from surya.model.layout.model import load_model as load_layout_model
|
| 20 |
+
from surya.model.layout.processor import load_processor as load_layout_processor
|
| 21 |
+
from surya.model.recognition.model import load_model as load_rec_model
|
| 22 |
+
from surya.model.recognition.processor import load_processor as load_rec_processor
|
| 23 |
+
from surya.model.table_rec.model import load_model as load_table_model
|
| 24 |
+
from surya.model.table_rec.processor import load_processor as load_table_processor
|
| 25 |
+
from surya.postprocessing.heatmap import draw_polys_on_image, draw_bboxes_on_image
|
| 26 |
+
from surya.ocr import run_ocr
|
| 27 |
+
from surya.postprocessing.text import draw_text_on_image
|
| 28 |
+
from PIL import Image
|
| 29 |
+
from surya.languages import CODE_TO_LANGUAGE
|
| 30 |
+
from surya.input.langs import replace_lang_with_code
|
| 31 |
+
from surya.schema import OCRResult, TextDetectionResult, LayoutResult, TableResult
|
| 32 |
+
from surya.settings import settings
|
| 33 |
+
from surya.tables import batch_table_recognition
|
| 34 |
+
from surya.postprocessing.util import rescale_bboxes, rescale_bbox
|
| 35 |
+
|
| 36 |
+
|
| 37 |
+
def load_det_cached():
|
| 38 |
+
return load_model(), load_processor()
|
| 39 |
+
|
| 40 |
+
def load_rec_cached():
|
| 41 |
+
return load_rec_model(), load_rec_processor()
|
| 42 |
+
|
| 43 |
+
def load_layout_cached():
|
| 44 |
+
return load_layout_model(), load_layout_processor()
|
| 45 |
+
|
| 46 |
+
def load_table_cached():
|
| 47 |
+
return load_table_model(), load_table_processor()
|
| 48 |
+
|
| 49 |
+
|
| 50 |
+
def text_detection(img) -> (Image.Image, TextDetectionResult):
|
| 51 |
+
pred = batch_text_detection([img], det_model, det_processor)[0]
|
| 52 |
+
polygons = [p.polygon for p in pred.bboxes]
|
| 53 |
+
det_img = draw_polys_on_image(polygons, img.copy())
|
| 54 |
+
return det_img, pred
|
| 55 |
+
|
| 56 |
+
|
| 57 |
+
def layout_detection(img) -> (Image.Image, LayoutResult):
|
| 58 |
+
pred = batch_layout_detection([img], layout_model, layout_processor)[0]
|
| 59 |
+
polygons = [p.polygon for p in pred.bboxes]
|
| 60 |
+
labels = [f"{p.label}-{p.position}" for p in pred.bboxes]
|
| 61 |
+
layout_img = draw_polys_on_image(polygons, img.copy(), labels=labels, label_font_size=18)
|
| 62 |
+
return layout_img, pred
|
| 63 |
+
|
| 64 |
+
|
| 65 |
+
def table_recognition(img, highres_img, filepath, page_idx: int, use_pdf_boxes: bool, skip_table_detection: bool) -> (Image.Image, List[TableResult]):
|
| 66 |
+
if skip_table_detection:
|
| 67 |
+
layout_tables = [(0, 0, highres_img.size[0], highres_img.size[1])]
|
| 68 |
+
table_imgs = [highres_img]
|
| 69 |
+
else:
|
| 70 |
+
_, layout_pred = layout_detection(img)
|
| 71 |
+
layout_tables_lowres = [l.bbox for l in layout_pred.bboxes if l.label == "Table"]
|
| 72 |
+
table_imgs = []
|
| 73 |
+
layout_tables = []
|
| 74 |
+
for tb in layout_tables_lowres:
|
| 75 |
+
highres_bbox = rescale_bbox(tb, img.size, highres_img.size)
|
| 76 |
+
table_imgs.append(
|
| 77 |
+
highres_img.crop(highres_bbox)
|
| 78 |
+
)
|
| 79 |
+
layout_tables.append(highres_bbox)
|
| 80 |
+
|
| 81 |
+
try:
|
| 82 |
+
page_text = get_page_text_lines(filepath, [page_idx], [highres_img.size])[0]
|
| 83 |
+
table_bboxes = get_table_blocks(layout_tables, page_text, highres_img.size)
|
| 84 |
+
except PdfiumError:
|
| 85 |
+
# This happens when we try to get text from an image
|
| 86 |
+
table_bboxes = [[] for _ in layout_tables]
|
| 87 |
+
|
| 88 |
+
if not use_pdf_boxes or any(len(tb) == 0 for tb in table_bboxes):
|
| 89 |
+
det_results = batch_text_detection(table_imgs, det_model, det_processor)
|
| 90 |
+
table_bboxes = [[{"bbox": tb.bbox, "text": None} for tb in det_result.bboxes] for det_result in det_results]
|
| 91 |
+
|
| 92 |
+
table_preds = batch_table_recognition(table_imgs, table_bboxes, table_model, table_processor)
|
| 93 |
+
table_img = img.copy()
|
| 94 |
+
|
| 95 |
+
for results, table_bbox in zip(table_preds, layout_tables):
|
| 96 |
+
adjusted_bboxes = []
|
| 97 |
+
labels = []
|
| 98 |
+
colors = []
|
| 99 |
+
|
| 100 |
+
for item in results.rows + results.cols:
|
| 101 |
+
adjusted_bboxes.append([
|
| 102 |
+
(item.bbox[0] + table_bbox[0]),
|
| 103 |
+
(item.bbox[1] + table_bbox[1]),
|
| 104 |
+
(item.bbox[2] + table_bbox[0]),
|
| 105 |
+
(item.bbox[3] + table_bbox[1])
|
| 106 |
+
])
|
| 107 |
+
labels.append(item.label)
|
| 108 |
+
if hasattr(item, "row_id"):
|
| 109 |
+
colors.append("blue")
|
| 110 |
+
else:
|
| 111 |
+
colors.append("red")
|
| 112 |
+
table_img = draw_bboxes_on_image(adjusted_bboxes, highres_img, labels=labels, label_font_size=18, color=colors)
|
| 113 |
+
return table_img, table_preds
|
| 114 |
+
|
| 115 |
+
def open_pdf(pdf_file):
|
| 116 |
+
return pypdfium2.PdfDocument(pdf_file)
|
| 117 |
+
|
| 118 |
+
def count_pdf(pdf_file):
|
| 119 |
+
doc = open_pdf(pdf_file)
|
| 120 |
+
return len(doc)
|
| 121 |
+
|
| 122 |
+
def get_page_image(pdf_file, page_num, dpi=96):
|
| 123 |
+
doc = open_pdf(pdf_file)
|
| 124 |
+
renderer = doc.render(
|
| 125 |
+
pypdfium2.PdfBitmap.to_pil,
|
| 126 |
+
page_indices=[page_num - 1],
|
| 127 |
+
scale=dpi / 72,
|
| 128 |
+
)
|
| 129 |
+
png = list(renderer)[0]
|
| 130 |
+
png_image = png.convert("RGB")
|
| 131 |
+
return png_image
|
| 132 |
+
|
| 133 |
+
def get_uploaded_image(in_file):
|
| 134 |
+
return Image.open(in_file).convert("RGB")
|
| 135 |
+
|
| 136 |
+
# Function for OCR
|
| 137 |
+
def ocr(img, highres_img, langs: List[str]) -> (Image.Image, OCRResult):
|
| 138 |
+
replace_lang_with_code(langs)
|
| 139 |
+
img_pred = run_ocr([img], [langs], det_model, det_processor, rec_model, rec_processor, highres_images=[highres_img])[0]
|
| 140 |
+
|
| 141 |
+
bboxes = [l.bbox for l in img_pred.text_lines]
|
| 142 |
+
text = [l.text for l in img_pred.text_lines]
|
| 143 |
+
rec_img = draw_text_on_image(bboxes, text, img.size, langs, has_math="_math" in langs)
|
| 144 |
+
return rec_img, img_pred
|
| 145 |
+
|
| 146 |
+
|
| 147 |
+
det_model, det_processor = load_det_cached()
|
| 148 |
+
rec_model, rec_processor = load_rec_cached()
|
| 149 |
+
layout_model, layout_processor = load_layout_cached()
|
| 150 |
+
table_model, table_processor = load_table_cached()
|
| 151 |
+
|
| 152 |
+
with gr.Blocks(title="Surya") as demo:
|
| 153 |
+
gr.Markdown("""
|
| 154 |
+
# Surya OCR Demo
|
| 155 |
+
|
| 156 |
+
This app will let you try surya, a multilingual OCR model. It supports text detection + layout analysis in any language, and text recognition in 90+ languages.
|
| 157 |
+
|
| 158 |
+
Notes:
|
| 159 |
+
- This works best on documents with printed text.
|
| 160 |
+
- Preprocessing the image (e.g. increasing contrast) can improve results.
|
| 161 |
+
- If OCR doesn't work, try changing the resolution of your image (increase if below 2048px width, otherwise decrease).
|
| 162 |
+
- This supports 90+ languages, see [here](https://github.com/VikParuchuri/surya/tree/master/surya/languages.py) for a full list.
|
| 163 |
+
|
| 164 |
+
Find the project [here](https://github.com/VikParuchuri/surya).
|
| 165 |
+
""")
|
| 166 |
+
|
| 167 |
+
with gr.Row():
|
| 168 |
+
with gr.Column():
|
| 169 |
+
in_file = gr.File(label="PDF file or image:", file_types=[".pdf", ".png", ".jpg", ".jpeg", ".gif", ".webp"])
|
| 170 |
+
in_num = gr.Slider(label="Page number", minimum=1, maximum=100, value=1, step=1)
|
| 171 |
+
in_img = gr.Image(label="Select page of Image", type="pil", sources=None)
|
| 172 |
+
|
| 173 |
+
text_det_btn = gr.Button("Run Text Detection")
|
| 174 |
+
layout_det_btn = gr.Button("Run Layout Analysis")
|
| 175 |
+
|
| 176 |
+
lang_dd = gr.Dropdown(label="Languages", choices=sorted(list(CODE_TO_LANGUAGE.values())), multiselect=True, max_choices=4, info="Select the languages in the image (if known) to improve OCR accuracy. Optional.")
|
| 177 |
+
text_rec_btn = gr.Button("Run OCR")
|
| 178 |
+
|
| 179 |
+
use_pdf_boxes_ckb = gr.Checkbox(label="Use PDF table boxes", value=True, info="Table recognition only: Use the bounding boxes from the PDF file vs text detection model.")
|
| 180 |
+
skip_table_detection_ckb = gr.Checkbox(label="Skip table detection", value=False, info="Table recognition only: Skip table detection and treat the whole image/page as a table.")
|
| 181 |
+
table_rec_btn = gr.Button("Run Table Rec")
|
| 182 |
+
with gr.Column():
|
| 183 |
+
result_img = gr.Image(label="Result image")
|
| 184 |
+
result_json = gr.JSON(label="Result json")
|
| 185 |
+
|
| 186 |
+
def show_image(file, num=1):
|
| 187 |
+
if file.endswith('.pdf'):
|
| 188 |
+
count = count_pdf(file)
|
| 189 |
+
img = get_page_image(file, num)
|
| 190 |
+
return [
|
| 191 |
+
gr.update(visible=True, maximum=count),
|
| 192 |
+
gr.update(value=img)]
|
| 193 |
+
else:
|
| 194 |
+
img = get_uploaded_image(file)
|
| 195 |
+
return [
|
| 196 |
+
gr.update(visible=False),
|
| 197 |
+
gr.update(value=img)]
|
| 198 |
+
|
| 199 |
+
in_file.upload(
|
| 200 |
+
fn=show_image,
|
| 201 |
+
inputs=[in_file],
|
| 202 |
+
outputs=[in_num, in_img],
|
| 203 |
+
)
|
| 204 |
+
in_num.change(
|
| 205 |
+
fn=show_image,
|
| 206 |
+
inputs=[in_file, in_num],
|
| 207 |
+
outputs=[in_num, in_img],
|
| 208 |
+
)
|
| 209 |
+
|
| 210 |
+
# Run Text Detection
|
| 211 |
+
def text_det_img(pil_image):
|
| 212 |
+
det_img, pred = text_detection(pil_image)
|
| 213 |
+
return det_img, pred.model_dump(exclude=["heatmap", "affinity_map"])
|
| 214 |
+
text_det_btn.click(
|
| 215 |
+
fn=text_det_img,
|
| 216 |
+
inputs=[in_img],
|
| 217 |
+
outputs=[result_img, result_json]
|
| 218 |
+
)
|
| 219 |
+
# Run layout
|
| 220 |
+
def layout_det_img(pil_image):
|
| 221 |
+
layout_img, pred = layout_detection(pil_image)
|
| 222 |
+
return layout_img, pred.model_dump(exclude=["segmentation_map"])
|
| 223 |
+
layout_det_btn.click(
|
| 224 |
+
fn=layout_det_img,
|
| 225 |
+
inputs=[in_img],
|
| 226 |
+
outputs=[result_img, result_json]
|
| 227 |
+
)
|
| 228 |
+
# Run OCR
|
| 229 |
+
def text_rec_img(pil_image, in_file, page_number, languages):
|
| 230 |
+
if in_file.endswith('.pdf'):
|
| 231 |
+
pil_image_highres = get_page_image(in_file, page_number, dpi=settings.IMAGE_DPI_HIGHRES)
|
| 232 |
+
else:
|
| 233 |
+
pil_image_highres = pil_image
|
| 234 |
+
rec_img, pred = ocr(pil_image, pil_image_highres, languages)
|
| 235 |
+
return rec_img, pred.model_dump()
|
| 236 |
+
text_rec_btn.click(
|
| 237 |
+
fn=text_rec_img,
|
| 238 |
+
inputs=[in_img, in_file, in_num, lang_dd],
|
| 239 |
+
outputs=[result_img, result_json]
|
| 240 |
+
)
|
| 241 |
+
def table_rec_img(pil_image, in_file, page_number, use_pdf_boxes, skip_table_detection):
|
| 242 |
+
if in_file.endswith('.pdf'):
|
| 243 |
+
pil_image_highres = get_page_image(in_file, page_number, dpi=settings.IMAGE_DPI_HIGHRES)
|
| 244 |
+
else:
|
| 245 |
+
pil_image_highres = pil_image
|
| 246 |
+
table_img, pred = table_recognition(pil_image, pil_image_highres, in_file, page_number - 1 if page_number else None, use_pdf_boxes, skip_table_detection)
|
| 247 |
+
return table_img, [p.model_dump() for p in pred]
|
| 248 |
+
table_rec_btn.click(
|
| 249 |
+
fn=table_rec_img,
|
| 250 |
+
inputs=[in_img, in_file, in_num, use_pdf_boxes_ckb, skip_table_detection_ckb],
|
| 251 |
+
outputs=[result_img, result_json]
|
| 252 |
+
)
|
| 253 |
+
|
| 254 |
+
demo.launch()
|
requirements.txt
ADDED
|
@@ -0,0 +1,4 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
torch==2.5.1
|
| 2 |
+
surya-ocr==0.7.0
|
| 3 |
+
gradio==5.8.0
|
| 4 |
+
huggingface-hub==0.26.3
|