Object Detection

Models:

• Yolo8n : Shashank Asthana's Model
• Yolo5_8 : Dev Bansal's Model
• LiveRCNN: Shreya Asthana's Model
• Yolov4 : Shreya Tripathi's Model

Streamlit Contributors:

• Tarandeep Singh Juneja

Helping Code

• text_to_audio (for output)
• activate (input)

My Model : Yolo8s_with_audio_activation

Integrated Yolov8 , NLP, Voice Commands

System Architecture:

---

1. Camera

   • Captures real-time images or video frames from the environment.
   • Sends the image stream to the next stage.

2. Object Detection Module

   • Processes the images using the YOLO model.
   • Detects and classifies objects within the frame.
   • Outputs object names, bounding boxes, and confidence scores.
   • Estimates whether objects are "closer" or "farther" based on the size of their bounding boxes.

3. Voice Command Module

   • Microphone Input: Listens to user commands using a speech recognition library (e.g., SpeechRecognition).
   • Processes commands like "activate," "stop," or "exit."

4. Text-to-Speech Module

   • Converts detected objects, their proximity (e.g., "closer" or "farther"), and scene descriptions into spoken feedback using Google Text-to-Speech (gTTS).
   • Plays the audio output through the speaker.

5. AI Response System

   • Combines object detection results, including proximity estimation (closer/farther), with voice input.
   • Provides intelligent responses to user commands and describes the detected scene (e.g., "The scene contains a person and a bicycle, both are closer").

6. User Interaction

   • Outputs include a live video stream with annotations (bounding boxes and labels), and audible scene descriptions indicating the proximity of objects.
   • Users interact through voice commands to control the system.

---

Flow of Data:
Camera → Object Detection Module (with proximity estimation) → AI Response System → (Text-to-Speech and Display) → User Interaction