Datasets:
First-pass object tracking on a driving-dataset-1 clip
Hi DrivingBench team,
I ran Ivy on a clip from driving-dataset-1 to see how it handles driving footage: https://drive.google.com/file/d/10Xf8LrxWtX2Hb-j_6kX1QeCHcnhkI0U5/view
To be upfront, it's a first pass. It tracks 5 traffic cones from about 30 to 40 seconds, following each one until the car passes it. It doesn't pick up new cones as they appear, though, so it's better suited to tracking known objects than to detecting new ones along the route.
I'm helping Interlatent (YC W27) get early users for Ivy. You tell your coding agent (Claude Code, Cursor or Codex) what to track, and it returns masks and tracks across the whole video.
Setup is the one line on interlatent.com, and new accounts get $10 in credits. A 10-second clip costs about 3 cents.
Curious whether something like this would be useful for driving data, or not.