app updated L202 tokens per chunk

app.py

@@ -31,22 +31,6 @@ palm.configure(api_key=api_key)
 yolo_pipe = pipeline("object-detection", model="hustvl/yolos-small")
 
 
-# Function to draw bounding boxes and labels on image
-def draw_boxes(image, predictions):
-    draw = ImageDraw.Draw(image)
-    font = ImageFont.load_default()
-
-    for pred in predictions:
-        label = pred["label"]
-        score = pred["score"]
-        box = pred["box"]
-        xmin, ymin, xmax, ymax = box.values()
-        draw.rectangle([xmin, ymin, xmax, ymax], outline="red", width=2)
-        draw.text((xmin, ymin), f"{label} ({score:.2f})", fill="red", font=font)
-
-    return image
-
-
 # Main function of the Streamlit app
 def main():
     st.title("Generative AI Demo on Camera Input/Image/PDF 💻")
@@ -215,7 +199,7 @@ def main():
         # Tokenize it
         st.warning("Start tokenzing ...")
         token_splitter = SentenceTransformersTokenTextSplitter(
-            chunk_overlap=0, tokens_per_chunk=256
+            chunk_overlap=0, tokens_per_chunk=20
         )
         token_split_texts = []
         for text in character_split_texts:
@@ -227,7 +211,7 @@ def main():
 
         # Generate a random string consisting of 10 uppercase letters and digits.
         random_string: str = "".join(
-            np.random.choice(list(string.ascii_uppercase + string.digits), size=10)
+            np.random.choice(list(string.ascii_uppercase + string.digits), size=20)
         )
 
         # Combine the random number and random string into one identifier.

utils/helpers.py

@@ -7,7 +7,7 @@ from typing import Any, Dict, List
 import pandas as pd
 import requests
 import streamlit as st
-from PIL import Image
+from PIL import Image, ImageDraw, ImageFont
 import google.generativeai as palm
 from pypdf import PdfReader
 from langchain.text_splitter import (
@@ -191,3 +191,42 @@ def displayPDF(file: str) -> None:
 
     # Using Streamlit to display the HTML embed string as unsafe HTML
     st.markdown(pdf_display, unsafe_allow_html=True)
+
+
+def draw_boxes(image: Any, predictions: List[Dict[str, Any]]) -> Any:
+    """
+    Draws bounding boxes and labels onto an image based on provided predictions.
+    
+    Parameters:
+    - image (Any): The image to annotate, which should support the PIL drawing interface.
+    - predictions (List[Dict[str, Any]]): A list of predictions where each prediction is a dictionary
+      containing 'label', 'score', and 'box' keys. The 'box' is another dictionary with 'xmin', 
+      'ymin', 'xmax', and 'ymax' as keys representing coordinates for the bounding box.
+    
+    Returns:
+    - Any: The annotated image with bounding boxes and labels drawn on it.
+    
+    Note:
+    - This function assumes that the incoming image supports the PIL ImageDraw interface.
+    - The function directly modifies the input image and returns it.
+    """
+    # Create a drawing context from the image
+    draw = ImageDraw.Draw(image)
+    # Load a default font for text drawing
+    font = ImageFont.load_default()
+
+    # Loop through all predictions and draw boxes with labels
+    for pred in predictions:
+        # Extracting label and score from the prediction
+        label = pred["label"]
+        score = pred["score"]
+        # Extracting the bounding box coordinates
+        box = pred["box"]
+        xmin, ymin, xmax, ymax = box.values()
+        # Draw a rectangle over the image using the box's coordinates
+        draw.rectangle([xmin, ymin, xmax, ymax], outline="red", width=2)
+        # Annotate the image with label and score at the top-left corner of the bounding box
+        draw.text((xmin, ymin), f"{label} ({score:.2f})", fill="red", font=font)
+
+    # Return the annotated image
+    return image