📷 Flutter Object Detection App

This is a Flutter-based real-time object detection app using TensorFlow Lite.
It processes live camera frames, provides real-time guidance to the user, and captures an image when the object is correctly positioned.

---

📌 Features

✔️ Real-time Object Detection using TensorFlow Lite
✔️ Live User Guidance ("Move Closer", "Move Farther", "Object in Position")
✔️ Automatic Image Capture when object is positioned correctly
✔️ Stores Image Metadata (date, time, object type)
✔️ GetX State Management for smooth UI updates ✔️ Performance Optimization by limiting frame processing

---

🚀 How to Run the App

1️⃣ Clone the Repository

git clone <your-repo-url>
cd <your-project-folder>

2️⃣ Install Dependencies

flutter pub get

3️⃣ Run the App on a Physical Device

flutter run

🛑 Important:

• This app requires a physical device as it uses the camera for object detection.
• Ensure you have TensorFlow Lite models in assets/models/.

---

📦 Dependencies Used

The following dependencies are used in this project (pubspec.yaml):

Package Name	Version	Purpose
flutter	latest	Core Flutter framework
get	^4.6.5	State management & navigation
camera	^0.10.5+2	Accessing device camera
tflite_flutter	^0.11.0	Running TensorFlow Lite models
image	^4.0.17	Image processing utilities
path_provider	^2.0.15	File system access
image_picker	^1.0.0	Selecting or capturing images
exif	^3.1.4	Extracting metadata from images
intl	^0.19.0	Formatting date/time

---

⚠️ Challenges Faced & Solutions

🧠 1. Performance Optimization (Device Overheating)

Issue:

• Continuous processing of camera frames caused the device to overheat and reduced performance.

Solution:

• Introduced frame skipping mechanism to reduce CPU/GPU workload.
• Added:

  int frameCounter = 0; // Used to limit FPS processing

• Optimized Image Processing: Frames are only processed every 5th frame, significantly improving performance.

---

🧠 2. Model Not Trained Well

Issue:

• The TensorFlow Lite model sometimes misclassifies objects or fails to detect them accurately.
• Inconsistent detection caused early or late object positioning messages.

Solution:

• Adjusted Confidence Threshold:
  We fine-tuned the model threshold (confidence = 0.5) to improve detection accuracy.

---

🧠 3. Sending Data Between Isolate and Main Thread

Issue:

• The TensorFlow Lite model runs in a separate isolate to avoid blocking the UI.
• However, sending camera frames from the main thread to the isolate caused serialization issues.
• CameraImage objects cannot be sent directly across isolates.

Solution:

• We used Background Isolate Channels to register the root isolate before running background operations.
• Implemented a Command-Based Messaging System:
  • The Main Isolate sends "detect" commands to the detector isolate.
  • The Detector Isolate processes the image and sends back "result" commands.
  • Captured Image Handling uses a capture lock (_isCapturing = true) to prevent multiple captures.

💪 This resolved the issue, allowing real-time image processing without UI lag!

🙌 Contributing

Want to improve the model accuracy or enhance the UI?
Feel free to fork the repo and submit a pull request!

git clone <your-repo-url>
git checkout -b feature-branch
git commit -m "Added new feature"
git push origin feature-branch

---

📝 License

This project is licensed under the MIT License.

---

🎯 Your project now has a structured README.md with all required details! 🚀
Let me know if you need any changes! 🚀💡

Object-Detection

Object-Detection