chatpaper / pdfextract_fun.py
zliang's picture
Update pdfextract_fun.py
3155ca8 verified
raw
history blame
6.67 kB
import warnings
from concurrent.futures import ThreadPoolExecutor, as_completed
import time
# Filter warnings about inputs not requiring gradients
warnings.filterwarnings("ignore", message="None of the inputs have requires_grad=True. Gradients will be None")
warnings.filterwarnings("ignore", message="torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument.")
import cv2
import os
import fitz # PyMuPDF
import numpy as np
import re
import pytesseract
import torch
from PIL import Image
from tqdm import tqdm
from unilm.dit.object_detection.ditod import add_vit_config
from detectron2.config import CfgNode as CN
from detectron2.config import get_cfg
from detectron2.utils.visualizer import ColorMode, Visualizer
from detectron2.data import MetadataCatalog
from detectron2.engine import DefaultPredictor
# Step 1: instantiate config
cfg = get_cfg()
add_vit_config(cfg)
cfg.merge_from_file("cascade_dit_base.yml")
# Step 2: add model weights URL to config
cfg.MODEL.WEIGHTS = "publaynet_dit-b_cascade.pth"
# Step 3: set device
cfg.MODEL.DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
#cfg.MODEL.DEVICE = "cuda"
# Step 4: define model
predictor = DefaultPredictor(cfg)
def analyze_image(img):
md = MetadataCatalog.get(cfg.DATASETS.TEST[0])
if cfg.DATASETS.TEST[0]=='icdar2019_test':
md.set(thing_classes=["table"])
else:
md.set(thing_classes=["text","title","list","table","figure"])
output = predictor(img)["instances"]
v = Visualizer(img[:, :, ::-1],
md,
scale=1.0,
instance_mode=ColorMode.SEGMENTATION)
result = v.draw_instance_predictions(output.to("cpu"))
result_image = result.get_image()[:, :, ::-1]
return result_image, output, v
def convert_pdf_to_jpg(pdf_path, output_folder, zoom_factor=2):
doc = fitz.open(pdf_path)
for page_num in range(len(doc)):
page = doc.load_page(page_num)
# Adjust zoom factor for higher resolution
mat = fitz.Matrix(zoom_factor, zoom_factor) # Create a Matrix with the zoom factor
pix = page.get_pixmap(matrix=mat) # Render the page using the matrix
output_file = f"{output_folder}/page_{page_num}.jpg"
pix.save(output_file)
def process_jpeg_images(output_folder):
for page_num in tqdm(range(len(os.listdir(output_folder))), desc="Processing the pdf"):
file_path = f"{output_folder}/page_{page_num}.jpg"
img = cv2.imread(file_path)
if img is None:
print(f"Failed to read {file_path}. Skipping.")
continue
result_image, output, v = analyze_image(img)
# Saving logic
save_extracted_instances(img, output, page_num,output_folder)
def save_extracted_instances(img, output, page_num, dest_folder, confidence_threshold=0.8):
class_names = {
0: "text",
1: "title",
2: "list",
3: "table",
4: "figure"
}
threshold_value = 0 # Standard deviation threshold
min_height = 0 # Minimum height threshold
instances = output.to("cpu")
boxes = instances.pred_boxes.tensor.numpy()
class_ids = instances.pred_classes.tolist()
scores = instances.scores.tolist() # Get prediction scores
image_counter = 1
for box, class_id, score in zip(boxes, class_ids, scores):
# Check if the prediction score meets the confidence threshold
if score >= confidence_threshold:
class_name = class_names.get(class_id, "unknown")
# Save only if class is 'figure' or 'table'
if class_name in ["figure", "table","text"]:
x1, y1, x2, y2 = map(int, box)
cropped_image = img[y1:y2, x1:x2]
if np.std(cropped_image) > threshold_value and (y2 - y1) > min_height:
save_path = os.path.join(dest_folder, f"page_{page_num}_{class_name}_{image_counter}.jpg")
cv2.imwrite(save_path, cropped_image)
image_counter += 1
def delete_files_in_folder(folder_path):
for filename in os.listdir(folder_path):
file_path = os.path.join(folder_path, filename)
if os.path.isfile(file_path):
os.remove(file_path)
def rename_files_sequentially(folder_path):
# Regex pattern to match 'page_{page_num}_{class_name}_{image_counter}.jpg'
pattern = re.compile(r'page_(\d+)_(\w+)_(\d+).jpg', re.IGNORECASE)
# List files in the folder
files = os.listdir(folder_path)
# Filter and sort files based on the regex pattern
sorted_files = sorted(
[f for f in files if pattern.match(f)],
key=lambda x: (int(pattern.match(x).group(1)), pattern.match(x).group(2).lower(), int(pattern.match(x).group(3)))
)
# Initialize an empty dictionary for counters
counters = {}
for filename in sorted_files:
match = pattern.match(filename)
if match:
page_num, class_name, _ = match.groups()
class_name = class_name.lower() # Convert class name to lowercase
# Initialize counter for this class if it doesn't exist
if class_name not in counters:
counters[class_name] = 1
# New filename format: '{class_name}_{sequential_number}.jpg'
new_filename = f"{class_name}_{counters[class_name]}.jpg"
counters[class_name] += 1
# Rename the file
os.rename(os.path.join(folder_path, filename), os.path.join(folder_path, new_filename))
#print(f"Renamed '{filename}' to '{new_filename}'")
def ocr_folder(folder_path):
# Regex pattern to match 'text_{number}.jpg'
pattern = re.compile(r'text_\d+\.jpg', re.IGNORECASE)
# Create a subfolder for the OCR text files
ocr_text_folder = os.path.join(folder_path, "ocr_results")
if not os.path.exists(ocr_text_folder):
os.makedirs(ocr_text_folder)
for filename in os.listdir(folder_path):
if pattern.match(filename):
image_path = os.path.join(folder_path, filename)
text = ocr_image(image_path)
# Save the OCR result to a text file in the subfolder
text_file_name = filename.replace('.jpg', '.txt')
text_file_path = os.path.join(ocr_text_folder, text_file_name)
with open(text_file_path, 'w') as file:
file.write(text)
#print(f"OCR result for {filename} saved to {text_file_path}\n")
def ocr_image(image_path):
image = Image.open(image_path)
text = pytesseract.image_to_string(image)
return text