updated for the real time fall detection

[shivam-khode01]

🚀 Real-Time Human Fall Detection using YOLOv8 & Pose Estimation
📌 Description:
This PR implements a real-time human fall detection system using the YOLOv8 model and pose estimation on a live webcam feed. It aims to detect falls based on body posture, track confidence levels, and trigger alerts accordingly.

✅ What's Done:
🔧 Integrated YOLOv8 with pose estimation for accurate detection of body keypoints.

📹 Used OpenCV to capture live video stream via webcam.

🧠 Fall detection logic based on:

Angle of the torso (vertical to horizontal transition)

Aspect ratio of bounding box (height vs width)

Y-axis displacement and speed of fall

🚨 Alert mechanism implemented with a red bounding box around the detected fall.

🛠️ Fixed bugs and errors, including:

Meshgrid warning from PyTorch

.grad warning on non-leaf tensors

ValueError: Type must be a sub-type of ndarray type — resolved by ensuring proper tensor-to-NumPy conversion.

🖼️ Sample Output:
Bounding boxes and keypoints on each person.

Fall alerts shown in red with label "Fall Detected".