README.md

metadata

title: Open-World Object Intelligence Platform
emoji: 👁️
colorFrom: indigo
colorTo: blue
sdk: gradio
sdk_version: 4.44.1
app_file: app.py
pinned: false
license: mit
hardware: gpu

Open-World Object Intelligence Platform

Complete System Architecture, Product Specification, ML Pipelines, Live Testing, Downloads, Guides, API, Database, and Hugging Face Deployment Plan

---

🌟 Key Features

1. Known Object Detection (RT-DETR): Fast, predictable object detection for standard COCO classes using RT-DETR.
2. Open-Vocabulary Discovery (YOLO-World): Zero-shot visual concept detection driven by dynamic text prompts without detector retraining.
3. Specific Object Recognition (Visual Embeddings): Deep PyTorch vision feature extractor + cosine similarity search to teach and recognize specific physical objects (e.g. My Cup vs generic cups).
4. 🔒 Lock Mode Target Filtering: Restricts output bounding boxes so that ONLY objects matching your lock target (e.g., 'My Cup', 'red backpack') are displayed on the live feed. Non-matching objects are completely suppressed.
5. Result Fusion Engine: Smart Non-Maximum Suppression (NMS) and IoU deduplication that prioritizes specific taught object identities over generic category labels.
6. Hugging Face Spaces & Gradio Web App: Sleek glassmorphic web interface built with Gradio and FastAPI ready for 1-click Hugging Face Spaces deployment.
7. Python SDK & OpenCV Examples: Clean client SDK (object_intelligence) with ready-to-run OpenCV webcam scripts.
8. Portable Custom Object Database: Portable SQLite database with ZIP import/export capabilities for taught object libraries.

---

🏗️ System Architecture

                 OBJECT INTELLIGENCE PLATFORM
                              |
      +-----------------------+-----------------------+
      |                       |                       |
      v                       v                       v
 KNOWN OBJECTS          OPEN VOCABULARY        SPECIFIC OBJECTS
    RT-DETR               YOLO-World             Embeddings
      |                       |                       |
      +-----------------------+-----------------------+
                              |
                              v
                        RESULT FUSION
                              |
                     +--------+--------+
                     |                 |
                     v                 v
                NORMAL MODE        LOCK MODE
                     |                 |
                     |           Target matching
                     |                 |
                     |                 v
                     |          Only target boxes
                     |                 |
                     +--------+--------+
                              |
                              v
                     OBJECT KNOWLEDGE DB

---

🚀 Quick Start

1. Installation

git clone https://github.com/user/Object-Intelligence.git
cd Object-Intelligence
pip install -r requirements.txt

2. Run Application Locally

python app.py

Open your browser at http://localhost:7860.

---

🐍 Python SDK Usage

from object_intelligence import ObjectDetector
import cv2

# Initialize detector client
detector = ObjectDetector()

# Select mode ('combined', 'known', 'open_vocabulary', 'specific')
detector.set_mode("combined")

# Enable Lock Mode for a specific target
detector.lock("My Cup")

# Process frame
frame = cv2.imread("sample.jpg")
annotated_frame, detections = detector.detect_and_draw(frame)

# Save result
cv2.imwrite("output.jpg", annotated_frame)

---

🧠 Teaching Custom Objects

Teach the system new physical objects using reference photos:

from object_intelligence import ObjectDetector

detector = ObjectDetector()
detector.add_object(
    name="My Cup",
    images=["cup_front.jpg", "cup_back.jpg", "cup_side.jpg"],
    category="Drinkware",
    description="Personal blue mug with handle"
)

---

🌐 Deploying to Hugging Face Spaces

1. Create a new Gradio Space on Hugging Face.
2. Upload this repository (app.py, requirements.txt, backend/, object_intelligence/, docs/, examples/).
3. Hugging Face Spaces will automatically install dependencies and run app.py!

---

📄 License

Distributed under the MIT License. See LICENSE for details.