Smart Camera

A smart camera application built with Flutter that focuses on real-time frame analysis, image quality enhancement, and on-device AI-based object detection.

The app is designed to intelligently process camera input and improve capture quality while running fully on the device without relying on external servers.

It consists of three main features:

Smart Frame Gate (Quality Filter):

Analyzes each camera frame in real time based on sharpness, motion stability, and exposure level, then automatically selects the best frames for capture.

Scan Mode:

A document-style image processing mode that enhances captured images through cropping, deskewing, contrast improvement, and adaptive thresholding to produce clean, scan-like results.

On-Device Object Detection (TFLite Integration):

Runs the EfficientDet-Lite0 model directly on the device for real-time object detection using TensorFlow Lite, with optimized preprocessing and coordinate mapping for accurate results.

Key Features

Real-time camera frame analysis with live performance metrics (FPS, processing time, accepted frames).

Fully on-device processing with no external API dependency.

Optimized performance using isolates for heavy computations.

Custom implementation of Non-Max Suppression (NMS) for object detection filtering.

Clean and simple UI with tab-based navigation between features.

Technologies Used

Flutter & Dart

Camera plugin

Image processing techniques

TensorFlow Lite (tflite_flutter)

Dart Isolates for performance optimization

Notes

The app runs in real time and is optimized for mid-range devices.

Performance was tested on a real device with stable results.

The project is structured for scalability and future feature expansion.