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




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
print("After parsing JSON from model response...")

  
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

def detect_and_draw_gradio(user_prompt: str, image: Image.Image, max_width: int = 1024):
    if image is None:
        return None, "Please upload an image."

    if not user_prompt or not user_prompt.strip():
        user_prompt = PROMPT

    image = image.convert("RGB")
    W, H = image.size

    # resize
    if W > max_width:
        newW = max_width
        newH = int(newW * H / W)
        im_resized = image.resize((newW, newH), Image.Resampling.LANCZOS)
    else:
        im_resized = image

    # send prompt + image
    response = model.generate_content([user_prompt, im_resized], generation_config=generation_config)
    raw_text = getattr(response, "text", "") or ""

    bounding_boxes = parse_json(raw_text)
    try:
        json.loads(bounding_boxes)
    except Exception:
        return im_resized, raw_text  # debugging

    out_img = plot_bounding_boxes(im_resized, bounding_boxes)
    return out_img, bounding_boxes