flow.yaml

_target_: rectified_point_flow.visualizer.FlowVisualizationCallback

save_dir: ${log_dir}

# We support the following renderers:
# - "mitsuba": Use Mitsuba for high quality ray-traced rendering.
# - "pytorch3d": Use PyTorch3D's PointRasterizer for fast rendering.
# - "none": Do not save any visualizations.
renderer: mitsuba

# Point cloud settings
center_points: false
scale_to_original_size: false

# Color settings
# We support the following colormaps:
# - "matplotlib:<name>": Use matplotlib colormap of the given name, e.g. "matplotlib:Set2"
# - "hue:<luminance>:<saturation>": Use hue-based color ring with the given luminance and saturation.
# - "shengyun": Use Shengyun's colormap defined in PyCG. We use this in the paper.
# - "default": Predefined 10 colors.
colormap: default

# Image settings
image_size: 480           # Output resolution (square image)
point_radius: 0.015       # Radius of sphere of each point

# Camera parameters
camera_dist: 4.0          # Distance from the camera to the point cloud center
camera_elev: 30.0         # Elevation angle above the horizon
camera_azim: 45.0         # Azimuth angle on the horizon plane
camera_fov: 40.0          # Field of view

# Trajectory GIF settings
save_trajectory: false    # Whether to save the trajectory as a GIF animation
trajectory_gif_fps: 25    # Frames per second
trajectory_gif_pause_last_frame: 1.0  # Pause for the last frame in seconds