Surya-OCR / app.py
artificialguybr's picture
Update app.py
9aaed47 verified
raw
history blame
3.76 kB
import gradio as gr
import json
from PIL import Image
# Assuming these imports work as expected, but you might need to adjust based on your actual package structure
from surya.ocr import run_ocr
from surya.detection import batch_detection
from surya.model.detection.segformer import load_model as load_det_model, load_processor as load_det_processor
from surya.model.recognition.model import load_model as load_rec_model
from surya.model.recognition.processor import load_processor as load_rec_processor
from surya.postprocessing.heatmap import draw_polys_on_image
# Load models and processors with print statements to confirm loading
print("Loading models and processors...")
det_model, det_processor = load_det_model(), load_det_processor()
rec_model, rec_processor = load_rec_model(), load_rec_processor()
print("Models and processors loaded successfully.")
# Load language codes
print("Loading language codes...")
with open("languages.json", "r") as file:
languages = json.load(file)
language_dict = {name: code for name, code in languages.items()}
print(f"Loaded languages: {list(language_dict.keys())}")
def ocr_function(img, lang_name):
print(f"OCR Function Called with lang_name: {lang_name}")
lang_code = language_dict[lang_name]
print(f"Language Code: {lang_code}")
# Ensure langs is a list of language codes, not a list of lists
predictions = run_ocr([img], [lang_code], det_model, det_processor, rec_model, rec_processor) # Corrected
print(f"Predictions: {predictions}")
if predictions:
img_with_text = draw_polys_on_image(predictions[0]["polys"], img)
return img_with_text, predictions[0]["text"]
else:
return img, "No text detected"
def text_line_detection_function(img):
print("Text Line Detection Function Called")
preds = batch_detection([img], det_model, det_processor)[0] # Assuming this returns a DetectionResult object
print(f"Detection Predictions: {preds}")
# Check if preds has an attribute 'bboxes' and use it
if hasattr(preds, 'bboxes'):
# Assuming draw_polys_on_image can work with the format of bboxes directly or you adapt it accordingly
img_with_lines = draw_polys_on_image([bbox.polygon for bbox in preds.bboxes], img)
return img_with_lines, preds
else:
raise AttributeError("DetectionResult object does not have 'bboxes' attribute")
with gr.Blocks() as app:
gr.Markdown("# Surya OCR and Text Line Detection")
with gr.Tab("OCR"):
with gr.Column():
ocr_input_image = gr.Image(label="Input Image for OCR", type="pil")
ocr_language_selector = gr.Dropdown(label="Select Language for OCR", choices=list(language_dict.keys()), value="English")
ocr_run_button = gr.Button("Run OCR")
with gr.Column():
ocr_output_image = gr.Image(label="OCR Output Image", type="pil", interactive=False)
ocr_text_output = gr.TextArea(label="Recognized Text")
ocr_run_button.click(fn=ocr_function, inputs=[ocr_input_image, ocr_language_selector], outputs=[ocr_output_image, ocr_text_output])
with gr.Tab("Text Line Detection"):
with gr.Column():
detection_input_image = gr.Image(label="Input Image for Detection", type="pil")
detection_run_button = gr.Button("Run Text Line Detection")
with gr.Column():
detection_output_image = gr.Image(label="Detection Output Image", type="pil", interactive=False)
detection_json_output = gr.JSON(label="Detection JSON Output")
detection_run_button.click(fn=text_line_detection_function, inputs=detection_input_image, outputs=[detection_output_image, detection_json_output])
if __name__ == "__main__":
app.launch()