app.py

import gradio as gr
import os
from io import BytesIO
from PIL import Image, ImageDraw, ImageFont
from PIL import ImageColor
import json
import google.generativeai as genai
from google.generativeai import types
from dotenv import load_dotenv
from IPython.display import display


# 1. SETUP API KEY
# ----------------
load_dotenv()
api_key = os.getenv("Gemini_API_Key") 
# Configure the Google AI library
genai.configure(api_key=api_key)


# 2. DEFINE MODEL AND INSTRUCTIONS

bounding_box_system_instructions = """
    Return bounding boxes as a JSON array with labels. Never return masks or code fencing. Limit to 25 objects.
    If an object is present multiple times, name them according to their unique characteristic (colors, size, position, unique characteristics, etc..).
      """
model = genai.GenerativeModel( model_name='gemini-2.5-flash', system_instruction=bounding_box_system_instructions)
generation_config = genai.types.GenerationConfig(
        temperature=0.5,
 
    )   



# 3. PREPARE IMAGE AND PROMPT
prompt = "Identify and label the objects in the image. Return only the JSON array of bounding boxes and labels as per the system instructions."
#image = "Images/cookies.jpg"  
#img = Image.open(BytesIO(open(image, "rb").read()))
# print(f"Original image size: {img.size}")

# resize the image to a max width of 1024 while maintaining aspect ratio
#im = Image.open(image).resize((1024, int(1024 * img.size[1] / img.size[0])), Image.Resampling.LANCZOS)
#print(f"Resized image size: {im.size}")
#im.show()


# Run model to find bounding boxes 
#response =  model.generate_content([prompt, im], generation_config=generation_config) 
 
#print(response.text)
# def generate_bounding_boxes(prompt, image):
    # response = model.generate_content([prompt, image], generation_config=generation_config)
    # return response.text


def parse_json(json_output): 
    lines = json_output.splitlines()
    for i, line in enumerate(lines):
        if line == "```json":
            json_output = "\n".join(lines[i+1:])  # Remove everything before "```json"
            json_output = json_output.split("```")[0]  # Remove everything after the closing "```"
            break
    return json_output

#bounding_boxes=parse_json(response.text)

#def plot_bounding_boxes(im, bounding_boxes):
    """
    Plots bounding boxes on an image with labels.
    """
    image = im.copy()
    draw = ImageDraw.Draw(image)
    font = ImageFont.load_default() 

    bounding_boxes_json = json.loads(bounding_boxes)
    for i, bounding_box in enumerate(bounding_boxes_json):
        print(f"Processing bounding box {i}: {bounding_box}")
        label = bounding_box["label"]
        x1, y1, x2, y2 = bounding_box["box_2d"]
        # Draw rectangle
        draw.rectangle(
            [(x1, y1), (x2, y2)],
            outline="red",
            width=10
        ) 
        # Draw label
        draw.text((x1 + 5, y1 + 5), label, fill="red", font=font) 
    return im
def plot_bounding_boxes(im, bounding_boxes):
    """
    Plots bounding boxes on an image with labels.
    """
    additional_colors = [colorname for (colorname, colorcode) in ImageColor.colormap.items()]

    im = im.copy()
    width, height = im.size
    draw = ImageDraw.Draw(im)
    colors = [
        'red', 'green', 'blue', 'yellow', 'orange', 'pink', 'purple', 'cyan',
        'lime', 'magenta', 'violet', 'gold', 'silver'
    ] + additional_colors

    try:
        # Use a default font if NotoSansCJK is not available
        try:
            font = ImageFont.load_default()
        except OSError:
            print("NotoSansCJK-Regular.ttc not found. Using default font.")
            font = ImageFont.load_default()

        bounding_boxes_json = json.loads(bounding_boxes)
        for i, bounding_box in enumerate(bounding_boxes_json):
            color = colors[i % len(colors)]
            abs_y1 = int(bounding_box["box_2d"][0] / 1000 * height)
            abs_x1 = int(bounding_box["box_2d"][1] / 1000 * width)
            abs_y2 = int(bounding_box["box_2d"][2] / 1000 * height)
            abs_x2 = int(bounding_box["box_2d"][3] / 1000 * width)

            if abs_x1 > abs_x2:
                abs_x1, abs_x2 = abs_x2, abs_x1

            if abs_y1 > abs_y2:
                abs_y1, abs_y2 = abs_y2, abs_y1

            # Draw bounding box and label
            draw.rectangle(((abs_x1, abs_y1), (abs_x2, abs_y2)), outline=color, width=4)
            if "label" in bounding_box:
                draw.text((abs_x1 + 8, abs_y1 + 6), bounding_box["label"], fill=color, font=font)
    except Exception as e:
        print(f"Error drawing bounding boxes: {e}")

    return im
#im_with_boxes = plot_bounding_boxes(im, bounding_boxes)
#display(im_with_boxes)
#im_with_boxes.save("output_imags/cookies_bounding_boxes.jpg")
#im_with_boxes.show()
#print("Bounding boxes plotted on image.")


def detect_objects(image , prompt):
    # Resize image
    image = image.resize((1024, int(1024 * image.size[1] / image.size[0])))

    # Generate bounding boxes
    response = model.generate_content([prompt, image], generation_config=generation_config)
    bounding_boxes = parse_json(response.text)

    # Draw boxes
    output_image = plot_bounding_boxes(image, bounding_boxes)

    return output_image, bounding_boxes

# ================== Gradio Interface ==================
interface = gr.Interface(
    fn=detect_objects,
    inputs=[gr.Image(type="pil"), gr.Textbox( label="Prompt", value="Identify and label the objects in the image. Return only the JSON array of bounding boxes.")],
    outputs=[gr.Image(label="Detected Objects"), gr.Textbox(label="Bounding Boxes JSON")],
    title="Object Detection with Gemini"
)

interface.launch(server_name="0.0.0.0", server_port=7860)