Cite arxiv in README body for paper linking

Link arxiv paper 2604.08626

Rename depth map label to Model Predicted Depth Map

Use brighter point colors with white outline for visibility

Keep only top-1 detection for Point and Box-to-Single-Object modes

Fix point drawing: scale radius with image size, remove black outline

Fix default image: convert to RGB to avoid RGBA issues

Fix white points/boxes: ensure uint8 before cv2 drawing

Update checkpoint to allenai/WildDet3D alldata v1.0

Remove repo link from description

Add depth input info to Space description

Update title to AI2 - WildDet3D

Add more tags for discoverability

Update Space metadata: add subtitle, Ai2 attribution, remove emoji

Remove Assumptions section from Terms of Use page

fix disclaimer wording: shape, without alignment

default text prompt: person.chair.monitor.pen

default text prompt: person.chair.monitor.plant

demo image: office with people, chairs, monitors, plants

text prompt: clearer label + match desk demo image

fix prompt colors: red box, green/red points with black outline

3D results: clean output without prompt box/points overlay

terms links: open in new tab with target=_blank

BEV labels: category name only, remove depth/size

default score threshold 0.3

default mode: Box-to-Single-Object with desk demo image

demo image: desk with laptop, plant, objects

depth confidence mask: app.py

depth confidence mask: wilddet3d/inference.py

depth confidence mask: wilddet3d/model.py

depth confidence mask: wilddet3d/data_types.py

default prompt: car.person.traffic light

demo image: NYC street scene

default prompt: person.horse.bus.bicycle

demo image: Indian street with people, animals, vehicles

default prompt: car.person.traffic light.building

demo image: NYC street scene with cars, people, traffic lights

outdoor demo: bus.car.street light.hot air balloon, remove intrinsics dep

remove intrinsics.npy (use default placeholder)

remove IoU threshold from disclaimer text

replace demo image with outdoor street scene

full-width layout, remove all orange borders and outlines

add Ai2 subtitle to main app page

update title to match paper: WildDet3D Scaling Promptable 3D Detection in the Wild, Ai2

remove orange border on Column after terms accept

add cross-category NMS (2D IoU > 0.8, keep higher combined score)

refine BEV disclaimer wording

add NMS note to BEV disclaimer

fix: indent all main app content inside main_app Column so terms gate works

add Terms of Use gate before main app