camera_hz float64 | duration_s float64 | id string | label string | messages dict | started_unix int64 | status string | summary string | version int64 | vision_raghouse-vision string | vision_thumbs list |
|---|---|---|---|---|---|---|---|---|---|---|
0 | 183.74111 | 0190 | {
"robot/camera/aux": 371,
"robot/camera/chest": 5566,
"robot/camera/waist": 5566,
"robot/camera/wrist_l": 5567,
"robot/camera/wrist_r": 5564,
"robot/command/ee/left": 0,
"robot/command/ee/right": 0,
"robot/command/joints/left": 0,
"robot/command/joints/right": 0,
"robot/control/owner": 184,
"robo... | 1,789,169,435 | success | #0190 black skirt -> B. bottoms | 1 | #0190 black skirt -> B. bottoms | [
"thumbs/0-aux.jpg",
"thumbs/1-chest.jpg"
] |
Bimanual Garment Sorting — Episode 0190
A single teleoperated episode of a bimanual robot picking a garment from a mixed laundry bin and sorting it to a labelled station.
Episode summary: black skirt → station B (bottoms)
Contents
| File | Description |
|---|---|
0190-chest-waist.mcap |
Chest and waist camera streams, JPEG frames in MCAP |
meta.json |
Original episode metadata as written by the recorder |
Recording
| Property | Value |
|---|---|
| Duration | 185.7 s (3 min 5.7 s) |
| Cameras | chest, waist |
| Frames | 11,132 (5,566 per camera) |
| Frame rate | 29.97 Hz per camera |
| Resolution | 1920 x 1080 |
| Size | 3.06 GB |
Frames are JPEG, carried as CameraImage messages in a custom alt-types
bincode schema. The MCAP container is zstd-compressed, though the payloads are
already compressed images so the container adds little.
Note on this release
This is a camera-only slice of a larger recording. The full episode also contains arm joint states, end-effector states, joint targets and teleop events at 10-100 Hz. Those non-camera channels total under 6 MB and can be published alongside on request.
The upstream episode additionally carried wrist and auxiliary camera streams, which are not included here.
Loading
from mcap.reader import make_reader
with open("0190-chest-waist.mcap", "rb") as f:
for schema, channel, message in make_reader(f).iter_messages():
# channel.topic is robot/camera/chest or robot/camera/waist
...
JPEG payloads can also be recovered directly by scanning for JPEG start and end markers, since each message wraps one complete frame.
- Downloads last month
- 83