[Dataset][ASE] Top Down View

[Astronirav]

Hi,

I downloaded ASE Dataset and I was wondering whether there is any way I can generate top-down view of the scene like it is shown in the promotional material:

I have a workaround where I am projecting semantic lables using pose and depth information and I'm getting something like this:

However, something shown on the website would be much better for my usecase.

[Tarunstk]

Hey @Astronirav I see. Let me check with the team and dataset owners if there is an easy way to do this and if there is a tool available internally.

The default view when viewing it is the front/first person view when visualising it right?

[Astronirav]

Yes, that is correct.

[Tarunstk]

Hey @Astronirav You can generate a top-down view directly from the ase_scene_language.txt using the code snippets that ship with projectaria_tools (projectaria_tools/projects/AriaSyntheticEnvironment/tutorial/code_snippets/).

Maybe try something like this

import plotly.graph_objects as go
from code_snippets.readers import read_language_file, read_trajectory_file
from code_snippets.interpreter import language_to_bboxes
from code_snippets.plotters import plot_box_wireframe, plot_trajectory

SCENE_PATH = "/path/to/ase_data/7"

# Load and interpret scene language into 3D boxes
entities = read_language_file(f"{SCENE_PATH}/ase_scene_language.txt")
boxes = language_to_bboxes(entities)

# Build plot traces
traces = [plot_box_wireframe(box) for box in boxes]

# Optionally add trajectory
trajectory = read_trajectory_file(f"{SCENE_PATH}/trajectory.csv")
traces.append(plot_trajectory(trajectory))

# Create figure with top-down orthographic camera
fig = go.Figure(data=traces)
fig.update_layout(
    template="plotly_dark",
    scene=dict(
        xaxis=dict(showticklabels=False, title=""),
        yaxis=dict(showticklabels=False, title=""),
        zaxis=dict(showticklabels=False, title=""),
        camera=dict(
            eye=dict(x=0, y=0, z=2),
            up=dict(x=0, y=1, z=0),
            projection=dict(type="orthographic"),
        ),
    ),
)
fig.show()