upgrade gradio to 4.26, better PDF report support

README.md

@@ -4,7 +4,7 @@ emoji: 📊
 colorFrom: pink
 colorTo: gray
 sdk: gradio
-sdk_version: 4.9.1
+sdk_version: 4.36.0
 app_file: app.py
 license: apache-2.0
 ---

app.py

@@ -185,7 +185,8 @@ examples = [
 def change_button_visibility(list_of_conponents):
     for comp in list_of_conponents:
         comp.value is not None
-    
+
+
 # add java script code for front right logo
 
 # Inputs
@@ -216,9 +217,10 @@ pitch_plot = gr.Plot(label="Pitch Contour and dB Analysis", show_label=False, co
 patient_name=gr.Textbox(placeholder="Patient Name", label="Patient Name", type="text")
 slp_name = gr.Textbox(placeholder="SLP Name", label="SLP Name", type="text")
 
-output_report = gr.File(interactive=False, type="filepath") 
+output_report = gr.File(interactive=False, type="filepath", visible=False) 
 # Report
-report_trigger = gr.Button(value="Download Report",  elem_classes="primary", variant="primary", visible=True)
+
+report_trigger = gr.DownloadButton(label="Download Report", value="report.pdf", elem_classes="primary", variant="primary")
 
 # Analysis
 run_all = gr.Button(value="Submit", elem_classes="primary", variant="primary")
@@ -228,6 +230,13 @@ with gr.Blocks(theme="KevinGeng/Laronix_ver2") as demo:
     #     gr.HTML(description, elem_classes="description")
     with gr.Row():
         with gr.Column(scale=1):
+            gr.Markdown('''
+                        # 0. Input User Information.
+                        ''')
+            with gr.Row():
+                patient_name.render()
+                slp_name.render()
+            
             gr.Markdown('''
                         # 1. Choose a sentence to read.
                         Copy & paste or type your sentence in the "Reference" field. \n
@@ -258,7 +267,7 @@ with gr.Blocks(theme="KevinGeng/Laronix_ver2") as demo:
             with gr.Row():
                 clear = gr.ClearButton(components=[ref, input_audio, exp_id], value="Clear All")
                 run_all.render()
-                run_all.click(fn=calc_mos, inputs=[exp_id, input_audio, ref, patient_name, slp_name], outputs=[nat_score, nat_plot, int_score, int_plot, hyp, phonemes, spk_rate, pitch_plot, output_report, Mean_Pitch, Mean_DB])
+                run_all.click(fn=calc_mos, inputs=[exp_id, input_audio, ref, patient_name, slp_name], outputs=[nat_score, nat_plot, int_score, int_plot, hyp, phonemes, spk_rate, pitch_plot, output_report, Mean_Pitch, Mean_DB], scroll_to_output=True)
                 # once run button is clicked, show nat_score, nat_plot, int_score, int_plot, these are invisible by default
     with gr.Row():
         gr.HTML("<hr>")
@@ -306,11 +315,12 @@ with gr.Blocks(theme="KevinGeng/Laronix_ver2") as demo:
             # Once the Submit button is clicked, nat_plot.value , int_plot.value are not None, the Download Report button will be visible
             
             # generate report
-            patient_name.render()
-            slp_name.render()
+            # patient_name.render()
+            # slp_name.render()
             output_report.render()
             report_trigger.render()
 
+
         
 demo.launch(share=True, allowed_paths=["local"])
             

local/pdf_generator.py

@@ -6,8 +6,12 @@ from reportlab.platypus import SimpleDocTemplate, Paragraph, Spacer, Image, Tabl
 from reportlab.lib.styles import getSampleStyleSheet, ParagraphStyle
 # from indicator_plot import Intelligibility_Plot, Naturalness_Plot
 # from pitch_contour import draw_spec_db_pitch
+from reportlab.pdfbase import pdfmetrics
+from reportlab.pdfbase.ttfonts import TTFont
+
 import io
 
+
 class MedicalReport:
     def __init__(self, patient_name, date, doctor_name, Nat_plot, Int_plot, ref, hyp, pitch_plot):
         
@@ -21,40 +25,59 @@ class MedicalReport:
         self.pitch_plot = Image(io.BytesIO(pitch_plot),  width=500, height=250)
 
     def generate_report(self, filename):
-        doc = SimpleDocTemplate(filename, pagesize=letter)
+        doc = SimpleDocTemplate(filename, pagesize=letter,  topMargin=0, bottomMargin=0)
+
         styles = getSampleStyleSheet()
-        right_logo = "./local/laronix_logo.jpg"  # Change this to the path of your logo image
+        # remove top margin
+        styles["Normal"].spaceBefore = 0.02 * inch
+        # Change the color of heading2 to green
+        styles["Heading2"].textColor = colors.green
+        # Change the color of heading2 to green but lighter, using RGB
+        # make heading3 50% transparent
+        styles["Heading3"].textColor.alpha = 0.7
 
+        # add under line to heading2
+        right_logo = "local/laronix-ai-logo-header.jpg"  # Change this to the path of your logo image
+        # load muli ttf and register it
+        pdfmetrics.registerFont(TTFont('Muli', 'local/Muli/Muli.ttf'))
+        
+        # change default font to muli
+        styles["Normal"].fontName = "Muli"
+        
         elements = []
-
         # Add logo on the right side
-        logo = Image(right_logo, width=100, height=35)
-        logo.hAlign = 'RIGHT'
+        # add header, logo width == page width
+        logo = Image(right_logo, width=600, height=85)
+        
+        # Align the logo in the center of the thw header
         elements.append(logo)
-        # add html header
         
-        # add split line
-        elements.append(Paragraph("<br></br>", styles["Normal"]))
-        # Add report content
-        elements.append(Paragraph("<b>Laronix SLP Report</b>", styles["Title"]))
-        elements.append(Paragraph(f"<b>Tester:</b> {self.patient_name}", styles["Normal"]))
-        elements.append(Paragraph(f"<b>SLP:</b> {self.doctor_name}", styles["Normal"]))
-        elements.append(Paragraph(f"<b>Date:</b> {self.date}", styles["Normal"]))
+        # add execute date on top right of the page
+        elements.append(Paragraph(f"Date: {self.date}", styles["BodyText"]))
+        
+        # Patient name, doctor name, and date
+        elements.append(Paragraph("User Information", styles["Heading2"]))
         
+        elements.append(Paragraph("<br></br>", styles["Normal"]))
+        elements.append(Paragraph("<br></br>", styles["Normal"]))
+        elements.append(Paragraph(f"<b>Participant:</b> {self.patient_name}, <b>\t    Instructor:</b> {self.doctor_name}", styles["Normal"]))
+
         elements.append(Paragraph("<br></br>", styles["Normal"]))
 
+        elements.append(Paragraph(f"<b>AI Analysis</b>", styles["Heading2"]))
+        elements.append(Paragraph("Speech Recognition", styles["Heading3"]))
         # Add ref and hyp
-        elements.append(Paragraph("<br></br><br/>", styles["Normal"]))  # Add some space
         elements.append(Paragraph(f"<b>Reference:</b> {self.ref}", styles["Normal"]))
         elements.append(Paragraph(f"<b>Hypothesis:</b> {self.hyp}", styles["Normal"]))
-        elements.append(Paragraph("<br></br><br/>", styles["Normal"]))
 
+
+        # AI Analysis
         # Create a table with two columns to organize the plots
         data = [
-            [Paragraph(f"<b>Naturalness Score:</b>", styles["Normal"]), Paragraph(f"<b>Intelligibility Score:</b>", styles["Normal"])],
+            [Paragraph(f"<b>Naturalness Score:</b>", styles["Heading3"]), Paragraph(f"<b>Intelligibility Score:</b>", styles["Heading3"])],
             [self.Nat_plot, self.Int_plot],
         ]
-        table = Table(data, colWidths=[200, 200])  # Adjust colWidths as needed
+        table = Table(data, colWidths=[230, 230])  # Adjust colWidths as needed
 
         # Add some space between the plots
         table.setStyle(TableStyle([('VALIGN', (0, 0), (-1, -1), 'TOP')]))
@@ -62,14 +85,14 @@ class MedicalReport:
 
         # Add the table to the elements list
         elements.append(table)
-        elements.append(Paragraph(f"<b>Pitch Plot:</b>", styles["Normal"]))
+        elements.append(Paragraph(f"<b>Acoustic Analysis</b>", styles["Heading2"]))
         elements.append(self.pitch_plot)
         # Build the PDF
         doc.build(elements)
         
         return filename
 
-# # Example usage:
+# Example usage:
 # patient_name = "John Doe"
 # AVA_history = 35
 # # Nat_plot = "./local/laronix_logo.jpg"
@@ -88,6 +111,148 @@ class MedicalReport:
 # pitch_plot = pitch_plot_buffer.read()
 # pitch_plot_buffer.close()
 
-# import pdb; pdb.set_trace()
+# # import pdb; pdb.set_trace()
 # report = MedicalReport(patient_name, date, doctor_name, Nat_plot, Int_plot, ref, hyp, pitch_plot)
-# x = report.generate_report("medical_report.pdf")
+# x = report.generate_report("medical_report.pdf")
+
+
+# from reportlab.pdfgen import canvas
+# from reportlab.lib.pagesizes import letter
+# from reportlab.lib.units import inch
+# from reportlab.pdfbase.ttfonts import TTFont
+# from reportlab.pdfbase import pdfmetrics
+# from reportlab.lib.pagesizes import letter
+# from reportlab.pdfgen import canvas
+# from reportlab.lib import colors
+# from reportlab.lib.units import inch
+# from reportlab.platypus import SimpleDocTemplate, Paragraph, Spacer, Image, Table, TableStyle
+# from reportlab.lib.styles import getSampleStyleSheet, ParagraphStyle
+# # from indicator_plot import Intelligibility_Plot, Naturalness_Plot
+# # from pitch_contour import draw_spec_db_pitch
+# import io
+
+
+# # Paths to the font files
+# poppins_font_path = 'local/Poppins/Poppins-Bold.ttf'  # Replace with the path to the Poppins font file
+# muli_font_path = 'local/Muli/Muli-Bold.ttf'  # Replace with the path to the Muli font file
+
+# # Register the fonts
+# pdfmetrics.registerFont(TTFont('Poppins', poppins_font_path))
+# pdfmetrics.registerFont(TTFont('Muli-Bold', muli_font_path))
+
+# # The path to the header image file
+# header_image_path = "/home/kevingeng/SSDisk/Laronix_voice_quality_checking_system_MICROPHONE/local/laronix_logo.jpg"
+
+# # Details for the content
+# user_name = "John Doe"
+# instructor_name = "Jane Smith"
+# testing_date = "2024-04-18"
+# spectrogtam_path = "003.jpg"
+
+# # Function to create a PDF with the new header and content
+# def create_pdf_with_header_and_content(output_path, header_path, user, instructor, date, spectrogram_path, mean_pitch, mean_decibel, ref, hyp):
+    
+#     c = canvas.Canvas(output_path, pagesize=letter)
+#     width, height = letter  # Get the dimensions of the letter size
+
+#     # Header image details
+#     # Make Header image width 3 inches
+    
+#     header_image_width = 3* inch
+#     header_image_height = 0.5 * inch
+
+#     # Calculate the position for the header
+#     x_position = (width - header_image_width) / 2
+#     y_position = height - header_image_height - 0.5 * inch  # 0.5 inch from the top of the page
+
+#     # Draw the header image
+#     c.drawImage(header_path, x_position, y_position, width=header_image_width, height=header_image_height, preserveAspectRatio=True)
+
+#     # make smaller padding for the header
+     
+#     # Title details
+#     title_text = "Voice Analysis Report"
+#     c.setFont("Poppins", 16)
+#     title_text_width = c.stringWidth(title_text, "Poppins", 16)
+#     title_x_position = (width - title_text_width) / 2
+#     title_y_position = y_position - 0.2 * inch  # 0.5 inch below the header image
+#     # Draw the title
+#     c.drawString(title_x_position, title_y_position, title_text)
+    
+#     # Content position details, move the content to up right next to the header
+#     name_x_position = x_position + header_image_width + 0.5 * inch
+#     content_x_position = 0.5 * inch
+#     content_y_position = y_position 
+    
+#     # User name details
+#     c.setFont("Muli-Bold", 12)
+#     c.drawString(name_x_position, content_y_position, f"User: {user}")
+#     # Instructor name details
+#     c.drawString(name_x_position, content_y_position - 0.3 * inch, f"Instructor: {instructor}")
+#     # Testing date details
+#     c.drawString(name_x_position, content_y_position - 0.6 * inch, f"Date: {date}")
+
+#     # # Draw a line to separate the header from the content
+#     # c.line(0, content_y_position - 0.8 * inch, width, content_y_position - 0.8 * inch)
+    
+#     # # Add mean pitch and decibel level in one line, still in the center of the page
+#     # c.setFont("Muli-Bold", 12)
+#     # c.drawString((width - 7 * inch) / 2, content_y_position - 1 * inch, f"Mean Pitch: {mean_pitch} Hz, Mean Decibel: {mean_decibel} dB")
+    
+#     # # Add reference and hypothesis in two line
+#     # c.setFont("Muli-Bold", 12)
+#     # c.drawString(content_x_position, content_y_position - 1.5 * inch, f"Reference: {ref}")
+#     # c.drawString(content_x_position, content_y_position - 2 * inch, f"Hypothesis: {hyp}")
+    
+#     # # Spectrogram and pitch contour details
+#     # spectrogram_image_height = 4 * inch  # Adjust as needed
+#     # # position the spectrogram image, under the split line, in the center of row
+#     # spectrogram_start_x_position = (width - 7* inch) / 2
+#     # spectrogram_start_y_position = content_y_position - 1 * inch - spectrogram_image_height
+    
+#     # # Draw the spectrogram image
+#     # c.drawImage(spectrogram_path, spectrogram_start_x_position, spectrogram_start_y_position, width=7* inch, height=spectrogram_image_height, preserveAspectRatio=True)
+
+#     doc = SimpleDocTemplate(output_path, pagesize=letter)
+#     styles = getSampleStyleSheet()
+
+#     elements = []
+
+#     # Add report content
+#     elements.append(Paragraph(f"<b>Tester:</b> {user}", styles["Normal"]))
+#     elements.append(Paragraph(f"<b>SLP:</b> {self}", styles["Normal"]))
+#     elements.append(Paragraph(f"<b>Date:</b> {self.date}", styles["Normal"]))
+    
+#     elements.append(Paragraph("<br></br>", styles["Normal"]))
+
+#     # Add ref and hyp
+#     elements.append(Paragraph("<br></br><br/>", styles["Normal"]))  # Add some space
+#     elements.append(Paragraph(f"<b>Reference:</b> {self.ref}", styles["Normal"]))
+#     elements.append(Paragraph(f"<b>Hypothesis:</b> {self.hyp}", styles["Normal"]))
+#     elements.append(Paragraph("<br></br><br/>", styles["Normal"]))
+
+#     # Create a table with two columns to organize the plots
+#     data = [
+#         [Paragraph(f"<b>Naturalness Score:</b>", styles["Normal"]), Paragraph(f"<b>Intelligibility Score:</b>", styles["Normal"])],
+#         [self.Nat_plot, self.Int_plot],
+#     ]
+#     table = Table(data, colWidths=[200, 200])  # Adjust colWidths as needed
+
+#     # Add some space between the plots
+#     table.setStyle(TableStyle([('VALIGN', (0, 0), (-1, -1), 'TOP')]))
+#     elements.append(Spacer(1, 12))
+
+#     # Add the table to the elements list
+#     elements.append(table)
+#     elements.append(Paragraph(f"<b>Pitch Plot:</b>", styles["Normal"]))
+#     elements.append(self.pitch_plot)
+#     # Build the PDF
+#     doc.build(elements)
+#     # Save the canvas
+#     c.save()
+
+# # Specify the path for the new PDF
+# new_pdf_path = '/home/kevingeng/SSDisk/Laronix_voice_quality_checking_system_MICROPHONE/report.pdf'
+
+# # Create the new PDF with the header and content
+# create_pdf_with_header_and_content(new_pdf_path, header_image_path, user_name, instructor_name, testing_date, spectrogtam_path, 100, 100, "ref", "hyp")

requirements.txt

@@ -1,4 +1,4 @@
-gradio==4.9.1
+gradio==4.26.0
 pytorch_lightning==1.5.9
 torch==1.12.1
 jiwer