realsense-sms-youtube

gistfile1.txt

import pyrealsense2 as rs
import numpy as np
import cv2
import requests
import http.client
import json

# Configure depth and color streams
pipeline = rs.pipeline()
config = rs.config()
config.enable_stream(rs.stream.depth, 640, 480, rs.format.z16, 30)
config.enable_stream(rs.stream.color, 640, 480, rs.format.bgr8, 30)

# Start streaming
pipeline.start(config)

# Load YOLOv3 model
net = cv2.dnn.readNetFromDarknet(
    'C:/Users/cmatthie/Projects/robotchat-tests/yolov3.cfg', 
    'C:/Users/cmatthie/Projects/robotchat-tests/yolov3.weights'
)
net.setPreferableBackend(cv2.dnn.DNN_BACKEND_OPENCV)
net.setPreferableTarget(cv2.dnn.DNN_TARGET_CPU)

# Load class labels
with open('C:/Users/cmatthie/Projects/robotchat-tests/coco.names', 'r') as f:
    classes = f.read().strip().split('\n')

person_detected = False

try:
    while True:
        # Wait for a coherent pair of frames: depth and color
        frames = pipeline.wait_for_frames()
        color_frame = frames.get_color_frame()
        depth_frame = frames.get_depth_frame()
        if not color_frame or not depth_frame:
            continue

        # Convert images to numpy arrays
        color_image = np.asanyarray(color_frame.get_data())

        # Prepare the image for person detection
        (h, w) = color_image.shape[:2]
        blob = cv2.dnn.blobFromImage(color_image, 1 / 255.0, (416, 416), swapRB=True, crop=False)
        net.setInput(blob)
        layer_names = net.getLayerNames()
        output_layers = [layer_names[i - 1] for i in net.getUnconnectedOutLayers()]
        detections = net.forward(output_layers)

        # Loop over the detections and show only one bounding box
        for output in detections:
            for detection in output:
                scores = detection[5:]
                class_id = np.argmax(scores)
                confidence = scores[class_id]

                # Filter out weak detections and ensure the detected object is a person
                if confidence > 0.5 and classes[class_id] == 'person':
                    box = detection[0:4] * np.array([w, h, w, h])
                    (centerX, centerY, width, height) = box.astype("int")
                    startX = int(centerX - (width / 2))
                    startY = int(centerY - (height / 2))

                    # Calculate the distance of the person from the camera
                    distance = depth_frame.get_distance(centerX, centerY)

                    # Draw the bounding box of the person along with the associated probability and distance
                    text = "{:.2f}%".format(confidence * 100)
                    y = startY - 10 if startY - 10 > 10 else startY + 10
                    cv2.rectangle(color_image, (startX, startY), (startX + width, startY + height),
                                  (0, 0, 255), 2)
                    cv2.putText(color_image, f"{text}, {distance:.2f}m", (startX, y),
                                cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 0, 255), 2)
                    
                    print(distance)
                    if not person_detected and distance < 1.0:
                        conn = http.client.HTTPSConnection("x.api.infobip.com")
                        payload = json.dumps({
                            "messages": [
                                {
                                    "destinations": [{"to":"14803813574"}],
                                    "from": "447491163443",
                                    "text": "a person was detected within 1 meter"
                                }
                            ]
                        })
                        headers = {
                            'Authorization': 'App 0fe535ba12a2e76df32d-x-2471fbb80a38',
                            'Content-Type': 'application/json',
                            'Accept': 'application/json'
                        }
                        conn.request("POST", "/sms/2/text/advanced", payload, headers)
                        res = conn.getresponse()
                        data = res.read()
                        print(data.decode("utf-8"))

                        person_detected = True
                    break
            break

        # Reset person_detected flag if no person is detected
        if not any(classes[np.argmax(detection[5:])] == 'person' and detection[5 + np.argmax(detection[5:])] > 0.5 for output in detections for detection in output):
            person_detected = False
            
        if distance > 1.0:
            person_detected = False

        # Show the output frame
        cv2.imshow("Frame", color_image)
        key = cv2.waitKey(1) & 0xFF

        # If the `q` key was pressed, break from the loop
        if key == ord("q"):
            break

finally:
    # Stop streaming
    pipeline.stop()
    cv2.destroyAllWindows()