gptmod.py

import numpy as np
import cv2
from keras.models import load_model

# Load pre-trained YOLO model and set classes
net = cv2.dnn.readNet("/home/mandarin/Documents/opencv/yolov3.cfg", "/home/mandarin/Documents/opencv/yolov3.weights")
classes = []
with open("/home/mandarin/Documents/opencv/coco.names", "r") as f:
    classes = [line.strip() for line in f.readlines()]

# Load pre-trained neural network for NLG
model = load_model("/path/to/pretrained/model.h5")

# Define function to extract objects from image
def extract_objects(image):
    # Preprocess image using YOLO
    blob = cv2.dnn.blobFromImage(image, 1 / 255.0, (416, 416), swapRB=True, crop=False)
    net.setInput(blob)
    outs = net.forward(net.getUnconnectedOutLayersNames())

    # Extract objects using YOLO
    objects = []
    for out in outs:
        for detection in out:
            scores = detection[5:]
            class_id = np.argmax(scores)
            confidence = scores[class_id]
            if confidence > 0.5:
                object_name = classes[class_id]
                objects.append((object_name, detection))

    return objects

# Define function to generate sentence based on detected objects
def generate_sentence(objects):
    # Create input vector for NLG model
    input_vec = np.zeros((1, len(objects), 5))
    for i, obj in enumerate(objects):
        class_id = classes.index(obj[0])
        x, y, w, h = obj[1][:4]
        input_vec[0][i] = [class_id, x, y, w, h]

    # Generate sentence using NLG model
    sentence = ""
    while "endseq" not in sentence:
        output = model.predict([input_vec, np.zeros((1, 1))])
        output_token = np.argmax(output[0, -1, :])
        word = classes[output_token]
        sentence += word + " "
        input_vec[0][len(sentence.split()) - 1][0] = output_token

    return sentence.strip()

# Load image and extract objects
image = cv2.imread("/path/to/image.jpg")
objects = extract_objects(image)

# Generate sentence based on objects
sentence = generate_sentence(objects)
print(sentence)