TDW Spatial Environment Experiment

A 3D spatial visualization experiment using ThreeDWorld (TDW) to capture multiple camera perspectives of a furnished room scene with field-of-view (FOV) visualization.

Overview

This project creates a virtual 3D environment containing furniture (sofa and chairs) and captures images from multiple camera positions and orientations. The experiment includes:

• Main Camera: Captures natural room views without visual markers
• Top-down Camera: Captures overhead views with FOV visualization markers on the ground plane
• FOV Visualization: Red markers showing the main camera's field of view area
• Multiple Perspectives: 10 different camera positions with 3 random orientations each (30 total image sets)

Requirements

• Python 3.7+
• ThreeDWorld (TDW) framework
• Required Python packages:
  • tdw
  • pathlib
  • math
  • random
  • os

Installation

1. Install TDW following the official documentation
2. Ensure TDW build is available and accessible
3. Install required dependencies:

   pip install tdw

Usage

Run the experiment:

python environment.py

The script will:

1. Create an empty room with furniture (sofa and two yellow chairs)
2. Set up two cameras (main view and top-down view)
3. Capture 30 image sets from different positions and orientations
4. Save all images to the specified output directory

Configuration

Key parameters you can modify:

• Output Directory: Change output_directory path in the main section
• Camera Positions: Modify the camera_positions list for different viewpoints
• FOV Settings: Adjust MAIN_CAMERA_FOV_ANGLE (default: 54.43°)
• Image Count: Change num_random_orientations_per_position for more/fewer views per position

Output

Images are saved to: C:\Users\Admin\Desktop\codebase\spatial\images\

Each image set includes:

• Main camera view: Natural room perspective without markers
• Top-down camera view: Overhead view with FOV visualization (red markers and blue camera position marker)

Image naming follows TDW's automatic naming convention with timestamps.

Scene Contents

• Room: 12x12 empty room
• Furniture:
  • Arflex strips sofa (center, ID: item_id)
  • Two yellow side chairs (positioned at angles, IDs: chair_1_id, chair_2_id)
• Cameras:
  • Main camera: 1.6m height, various positions
  • Top-down camera: 10m height, looking down at scene center

Features

• FOV Visualization: Real-time field-of-view markers on ground plane
• Random Orientations: Each camera position includes 3 random look-at targets
• Dual Perspective Capture: Simultaneous main view and overhead analysis view
• Automatic Cleanup: FOV markers are removed after each capture
• Progress Tracking: Console output shows capture progress

Technical Details

• Image Resolution: 512x512 pixels
• Camera FOV: 54.43° (matches realistic camera settings)
• FOV Markers: Small red spheres filling the camera's view area
• Ground Plane: Y=0.01 to ensure markers are visible above ground

Troubleshooting

• Ensure TDW build is running and accessible on the specified port (default: 1071)
• Check that output directory has write permissions
• Verify all required TDW model libraries are available (models_core.json, models_flex.json)