Browser-native 3D perception explorer for Waymo Open Dataset v2.0 Perception
No install. No server. Your data never leaves your browser.
Waymo Open Dataset is one of the richest public self-driving datasets out there, but actually looking at the data is painful. The official tools need Python + TensorFlow. Jupyter gives you static plots. Foxglove is paid. And you can't even use the raw Parquet files without preprocessing scripts.
You just want to drop in the files and see what's inside.
- See what the car sees: explore real self-driving scenes in 3D with LiDAR point clouds and 5 synchronized camera views
- 3D object models: vehicles, pedestrians, and cyclists rendered as 3D models with color-coded tracking
- Camera POV mode: click a camera to jump into its viewpoint in 3D, compare what the sensor sees side by side
- Cross-modal linking: hover over a camera detection and its 3D counterpart lights up, and vice versa
![]() |
![]() |
pov.mp4
Already have Waymo Open Dataset v2.0 Perception downloaded? Just open the live demo, drop your files, and go.
Don't have the data yet? It's free with a Google account.
Download script
# Install Google Cloud CLI: https://cloud.google.com/sdk/docs/install
gcloud auth login
BUCKET="gs://waymo_open_dataset_v_2_0_1/training"
COMPONENTS="vehicle_pose lidar_calibration camera_calibration lidar_box lidar lidar_camera_projection camera_image"
N=1 # Number of segments to download (~500 MB each)
SEGMENTS=$(gsutil ls "$BUCKET/vehicle_pose/*.parquet" | head -$N | xargs -I{} basename {} .parquet)
for SEG in $SEGMENTS; do
echo "Downloading $SEG"
for C in $COMPONENTS; do
mkdir -p waymo_data/$C
gsutil -m cp "$BUCKET/$C/$SEG.parquet" "waymo_data/$C/"
done
doneThen drag & drop the waymo_data/ folder into the app.
git clone https://github.com/heejaekim/waymo-perception-studio.git
cd waymo-perception-studio
npm install
npm run devReact 19 · TypeScript · Three.js · R3F · Vite · Zustand · hyparquet · Web Workers
Chrome / Edge recommended (folder drag & drop + folder picker). Firefox / Safari support folder drag & drop only.
MIT



