dino.txt

import gradio as gr
import cv2
import numpy as np
import tempfile
import os
import requests
from dds_cloudapi_sdk import Config, Client
from dds_cloudapi_sdk.tasks.dinox import DinoxTask
from dds_cloudapi_sdk import TextPrompt
from dds_cloudapi_sdk.tasks.types import DetectionTarget
from roboflow import Roboflow
from dotenv import load_dotenv

# ========== Konfigurasi ========== 
load_dotenv()

# Roboflow Config
rf_api_key = os.getenv("ROBOFLOW_API_KEY")
workspace = os.getenv("ROBOFLOW_WORKSPACE")
project_name = os.getenv("ROBOFLOW_PROJECT")
model_version = int(os.getenv("ROBOFLOW_MODEL_VERSION"))

# DINO-X Config
DINOX_API_KEY = os.getenv("DINO_X_API_KEY")
DINOX_PROMPT = "beverage . bottle . cans . boxed milk . milk"

# Inisialisasi Model YOLO (Roboflow)
rf = Roboflow(api_key=rf_api_key)
project = rf.workspace(workspace).project(project_name)
yolo_model = project.version(model_version).model

# Inisialisasi DINO-X API Client
dinox_config = Config(DINOX_API_KEY)
dinox_client = Client(dinox_config)

# Fungsi untuk mendeteksi objek pada gambar dan video
def detect_combined(image_path_or_video_path, is_video=False):
    # Jika input adalah video
    if is_video:
        return detect_objects_in_video(image_path_or_video_path)
    
    # Jika input adalah gambar
    return detect_objects_in_image(image_path_or_video_path)

def detect_objects_in_image(image_path):
    try:
        # Membaca gambar
        img = cv2.imread(image_path)

        # --- Deteksi menggunakan YOLO (Nestlé) ---
        yolo_pred = yolo_model.predict(image_path, confidence=50, overlap=80).json()

        # Hitung produk Nestlé per kelas
        nestle_class_count = {}
        nestle_boxes = []
        for pred in yolo_pred['predictions']:
            class_name = pred['class']
            nestle_class_count[class_name] = nestle_class_count.get(class_name, 0) + 1
            nestle_boxes.append((pred['x'], pred['y'], pred['width'], pred['height']))

        # --- Deteksi menggunakan DINO-X (Unclassified Products) ---
        image_url = dinox_client.upload_file(image_path)
        task = DinoxTask(
            image_url=image_url,
            prompts=[TextPrompt(text=DINOX_PROMPT)],
            bbox_threshold=0.25,
            targets=[DetectionTarget.BBox]
        )
        dinox_client.run_task(task)
        dinox_pred = task.result.objects

        # Hitung produk kompetitor yang tidak tumpang tindih dengan deteksi YOLO
        competitor_class_count = {}
        competitor_boxes = []
        for obj in dinox_pred:
            dinox_box = obj.bbox
            # Filter objek yang sudah terdeteksi oleh YOLO (Overlap detection)
            if not is_overlap(dinox_box, nestle_boxes):  # Ignore if overlap with YOLO detections
                class_name = obj.category.strip().lower()
                competitor_class_count[class_name] = competitor_class_count.get(class_name, 0) + 1
                competitor_boxes.append({
                    "class": class_name,
                    "box": dinox_box,
                    "confidence": obj.score
                })

        # --- Overlay Teks untuk Total Produk ---
        nestle_count_text = ""
        total_nestle = 0
        for class_name, count in nestle_class_count.items():
            nestle_count_text += f"{class_name}: {count}\n"
            total_nestle += count
        nestle_count_text += f"\nTotal Nestlé Products: {total_nestle}"

        unclassified_count_text = ""
        total_unclassified = 0
        for class_name, count in competitor_class_count.items():
            unclassified_count_text += f"{class_name}: {count}\n"
            total_unclassified += count
        unclassified_count_text += f"\nTotal Unclassified Products: {total_unclassified}"

        # --- Visualisasi Deteksi YOLO (Nestlé) ---
        for pred in yolo_pred['predictions']:
            x, y, w, h = pred['x'], pred['y'], pred['width'], pred['height']
            cv2.rectangle(img, (int(x-w/2), int(y-h/2)), (int(x+w/2), int(y+h/2)), (0,255,0), 2)
            cv2.putText(img, pred['class'], (int(x-w/2), int(y-h/2-10)), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0,255,0), 2)

        # --- Visualisasi Deteksi DINO-X (Unclassified) ---
        for comp in competitor_boxes:
            x1, y1, x2, y2 = comp['box']
            display_name = "unclassified"
            cv2.rectangle(img, (int(x1), int(y1)), (int(x2), int(y2)), (0, 0, 255), 2)
            cv2.putText(img, f"{display_name} {comp['confidence']:.2f}",
                        (int(x1), int(y1-10)), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 0, 255), 2)

        # Simpan gambar output
        output_path = "/tmp/combined_output_image.jpg"
        cv2.imwrite(output_path, img)

        return output_path, nestle_count_text + "\n" + unclassified_count_text

    except Exception as e:
        return image_path, f"Error: {str(e)}"

def detect_objects_in_video(video_path):
    temp_output_path = "/tmp/output_video.mp4"
    temp_frames_dir = tempfile.mkdtemp()
    frame_count = 0
    previous_detections = {}  # Untuk menyimpan deteksi objek dari frame sebelumnya

    # Membuka video
    video = cv2.VideoCapture(video_path)
    frame_rate = int(video.get(cv2.CAP_PROP_FPS))
    frame_width = int(video.get(cv2.CAP_PROP_FRAME_WIDTH))
    frame_height = int(video.get(cv2.CAP_PROP_FRAME_HEIGHT))
    frame_size = (frame_width, frame_height)

    # VideoWriter untuk menyimpan hasil video
    fourcc = cv2.VideoWriter_fourcc(*'mp4v')
    output_video = cv2.VideoWriter(temp_output_path, fourcc, frame_rate, frame_size)

    while True:
        ret, frame = video.read()
        if not ret:
            break

        # Simpan frame sementara untuk prediksi
        frame_path = os.path.join(temp_frames_dir, f"frame_{frame_count}.jpg")
        cv2.imwrite(frame_path, frame)

        # --- Deteksi menggunakan YOLO (Nestlé) ---
        yolo_pred = yolo_model.predict(frame_path, confidence=50, overlap=80).json()

        # Hitung produk Nestlé per kelas
        nestle_class_count = {}
        nestle_boxes = []
        for pred in yolo_pred['predictions']:
            class_name = pred['class']
            nestle_class_count[class_name] = nestle_class_count.get(class_name, 0) + 1
            nestle_boxes.append((pred['x'], pred['y'], pred['width'], pred['height']))

        # --- Deteksi menggunakan DINO-X (Unclassified Products) ---
        image_url = dinox_client.upload_file(frame_path)
        task = DinoxTask(
            image_url=image_url,
            prompts=[TextPrompt(text=DINOX_PROMPT)],
            bbox_threshold=0.25,
            targets=[DetectionTarget.BBox]
        )
        dinox_client.run_task(task)
        dinox_pred = task.result.objects

        # Hitung produk kompetitor yang tidak tumpang tindih dengan deteksi YOLO
        competitor_class_count = {}
        competitor_boxes = []
        for obj in dinox_pred:
            dinox_box = obj.bbox
            # Filter objek yang sudah terdeteksi oleh YOLO (Overlap detection)
            if not is_overlap(dinox_box, nestle_boxes):  # Ignore if overlap with YOLO detections
                class_name = obj.category.strip().lower()
                competitor_class_count[class_name] = competitor_class_count.get(class_name, 0) + 1
                competitor_boxes.append({
                    "class": class_name,
                    "box": dinox_box,
                    "confidence": obj.score
                })

        # --- Overlay Teks untuk Total Produk ---
        nestle_count_text = ""
        total_nestle = 0
        for class_name, count in nestle_class_count.items():
            nestle_count_text += f"{class_name}: {count}\n"
            total_nestle += count
        nestle_count_text += f"\nTotal Nestlé Products: {total_nestle}"

        unclassified_count_text = ""
        total_unclassified = 0
        for class_name, count in competitor_class_count.items():
            unclassified_count_text += f"{class_name}: {count}\n"
            total_unclassified += count
        unclassified_count_text += f"\nTotal Unclassified Products: {total_unclassified}"

        # Overlay teks ke frame
        y_offset = 20
        for line in nestle_count_text.split("\n"):
            cv2.putText(frame, line, (10, y_offset), cv2.FONT_HERSHEY_SIMPLEX, 0.6, (255, 255, 255), 2)
            y_offset += 30

        y_offset += 30  # Slight gap between sections
        for line in unclassified_count_text.split("\n"):
            cv2.putText(frame, line, (10, y_offset), cv2.FONT_HERSHEY_SIMPLEX, 0.6, (255, 255, 255), 2)
            y_offset += 30

        # --- Visualisasi Deteksi YOLO (Nestlé) ---
        for pred in yolo_pred['predictions']:
            x, y, w, h = pred['x'], pred['y'], pred['width'], pred['height']
            cv2.rectangle(frame, (int(x-w/2), int(y-h/2)), (int(x+w/2), int(y+h/2)), (0,255,0), 2)
            cv2.putText(frame, pred['class'], (int(x-w/2), int(y-h/2-10)), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0,255,0), 2)

        # --- Visualisasi Deteksi DINO-X (Unclassified) ---
        for comp in competitor_boxes:
            x1, y1, x2, y2 = comp['box']
            display_name = "unclassified"
            cv2.rectangle(frame, (int(x1), int(y1)), (int(x2), int(y2)), (0, 0, 255), 2)
            cv2.putText(frame, f"{display_name} {comp['confidence']:.2f}",
                        (int(x1), int(y1-10)), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 0, 255), 2)

        # Tulis frame ke video output
        output_video.write(frame)
        frame_count += 1

    video.release()
    output_video.release()

    return temp_output_path

def is_overlap(box1, boxes2, threshold=0.3):
    # Fungsi untuk deteksi overlap bounding box
    x1_min, y1_min, x1_max, y1_max = box1
    for b2 in boxes2:
        x2, y2, w2, h2 = b2
        x2_min = x2 - w2/2
        x2_max = x2 + w2/2
        y2_min = y2 - h2/2
        y2_max = y2 + h2/2

        # Hitung area overlap
        dx = min(x1_max, x2_max) - max(x1_min, x2_min)
        dy = min(y1_max, y2_max) - max(y1_min, y2_min)
        if (dx >= 0) and (dy >= 0):
            area_overlap = dx * dy
            area_box1 = (x1_max - x1_min) * (y1_max - y1_min)
            if area_overlap / area_box1 > threshold:
                return True
    return False

# ========== Gradio Interface ========== 
with gr.Blocks(theme=gr.themes.Base(primary_hue="teal", secondary_hue="teal", neutral_hue="slate")) as iface:
    gr.Markdown("""<div style="text-align: center;"><h1>NESTLE - STOCK COUNTING</h1></div>""")
    
    with gr.Row():
        with gr.Column():
            input_image = gr.Image(type="pil", label="Input Image")
            detect_image_button = gr.Button("Detect Image")
            output_image = gr.Image(label="Detect Object")
            output_text = gr.Textbox(label="Counting Object")
            detect_image_button.click(fn=detect_combined, inputs=input_image, outputs=[output_image, output_text])

        with gr.Column():
            input_video = gr.Video(label="Input Video")
            detect_video_button = gr.Button("Detect Video")
            output_video = gr.Video(label="Output Video")
            detect_video_button.click(fn=detect_objects_in_video, inputs=input_video, outputs=[output_video])

iface.launch()