---
title: VisionAI - Object Detection & Human Pose Estimation
emoji: 🤖
colorFrom: blue
colorTo: indigo
sdk: gradio
sdk_version: 5.33.0
app_file: app.py
pinned: false
license: mit
tags:
  - object-detection
  - pose-estimation
  - weapon-detection
  - threat-analysis
  - yolo
  - computer-vision
short_description: YOLO11 Object Detection, Pose Estimation & Threat Analysis
---

# 🤖 VisionAI — Object Detection, Human Pose Estimation & Threat Analysis

**Semester Project** — YOLO11-powered computer vision with real-time pose threat classification.

## Models Included
| Model | File | Purpose |
|-------|------|---------|
| Object Detection | `yolo11m.pt` | 80+ COCO classes |
| Pose Estimation | `yolo11m-pose.pt` | 17 body keypoints |
| Segmentation | `yolo11m-seg.pt` | Instance masks |
| Classification | `yolo11m-cls.pt` | ImageNet-1k |
| OBB Detection | `yolo11m-obb.pt` | Rotated bounding boxes |
| **Weapon Detection** | `weapon_detection.pt` | **Custom bundled model ★** |

## 🎯 Pose Threat Analysis
Every detected person's posture is analysed using 17 COCO keypoints and classified:

| Status | Criteria |
|--------|----------|
| 🟢 **NORMAL** | Relaxed / standing / walking posture |
| 🟡 **SUSPICIOUS** | Crouching / leaning / unusual body angle |
| 🔴 **THREATENING** | Raised arms / aggressive stance / weapon in frame |

Works on **images**, **videos** (FPS-based scanning), and **live webcam**.

## Video FPS-Based Scanning
| Scan FPS | Interval at 30fps | Speed |
|----------|-------------------|-------|
| 1 fps | every 30th frame | Fastest |
| 5 fps | every 6th frame | Balanced ✅ |
| 15 fps | every 2nd frame | Thorough |
| 24 fps | every frame | Full quality |

## Deploy (3 files only)
Upload to a HuggingFace Gradio Space:
- `app.py`
- `requirements.txt`
- `weapon_detection.pt`
- `README.md`