Sightsense is a real-time computer vision tool designed to assist visually impaired users by recognizing objects, reading text, and delivering spoken feedback. It combines optical character recognition (OCR), an object detection model, and a text-to-speech engine to transform live camera input into meaningful audio cues.

Core Components

• Camera Input
  Uses a webcam or device camera to capture video frames. The default implementation leverages OpenCV to access and read frames from the camera feed.
• Object Detection
  A pretrained model, configured within the repository, processes each captured frame and draws bounding boxes or labels around recognized items (for example, YOLO or TensorFlow-based models).
• Text Recognition
  Integrates an OCR library (such as Tesseract) to parse text within the frame and return recognized strings.
• Text-to-Speech Engine
  Converts the recognized labels and text into audio output. Depending on the user’s environment, the code might leverage libraries like pyttsx3, gTTS, or a cloud TTS API.

1. Frame Acquisition
   • Continuously reads camera frames.
   • Passes raw images to the processing pipeline.
2. Processing Pipeline
   • Object Detection: Processes each frame, identifies objects, and returns detection results (class labels, confidence scores, bounding box coordinates).
   • OCR Module (Conditional): If enabled, crops relevant regions containing text and applies OCR to extract text data.
3. Output Generation
   • Merges object detection results (and OCR findings if available) into a textual description.
   • Uses a text-to-speech engine to generate human-audible feedback.
4. Real-Time Feedback Loop
   • Plays the generated audio description.
   • Continues acquiring subsequent frames.

1. Clone the repository.
2. Install required Python packages: