text
stringlengths
0
133
# Edged USLAM Dataset – Davis346 UAV Sequences
This dataset was collected using a DAVIS346 event camera mounted on a UAV platform.
The recordings include synchronized events, frames, and IMU data, along with ground-truth poses from a motion capture system (Vicon).
--------------------------------------------------
1. Sensor Topics
--------------------------------------------------
- Event camera (DAVIS346):
• /dvs/events → asynchronous event stream
• /dvs/image_raw → grayscale intensity frames
• /dvs/imu → inertial measurements (accelerometer + gyroscope)
- Ground truth (Vicon):
• /mavros/vision_pose/pose (used in motion/line.bag, motion/square.bag, motion/turner.bag)
• /local_pose_vicon/pose (used in filtered/ and unfiltered/ sequences)
--------------------------------------------------
2. Dataset Structure
--------------------------------------------------
- motion/
- line.bag
- square.bag
- turner.bag
→ Ground truth: /mavros/vision_pose/pose
- filtered/
→ IR-filtered lens attached to the DAVIS346
→ Ground truth: /local_pose_vicon/pose
- unfiltered/
→ Default (non-filtered) lens on the DAVIS346
→ Ground truth: /local_pose_vicon/pose
--------------------------------------------------
3. Notes
--------------------------------------------------
- All sequences are timestamp-synchronized across topics.
- The IR filter (filtered/) improves robustness under strong illumination but reduces sensitivity.
- The unfiltered/ data preserve the original DAVIS346 response, including IR wavelengths.
- Ground truth in both cases provides 6-DoF UAV pose in the Vicon frame.
--------------------------------------------------
4. Usage
--------------------------------------------------
When training or evaluating algorithms, ensure that the correct ground-truth topic is selected depending on the sequence folder.
# DAVIS346 Calibration Parameters (DVS_ext.yaml Explained)
This file contains the intrinsic, distortion, and extrinsic calibration parameters
for the DAVIS346 event camera (DAVIS-IJRR17 profile).
--------------------------------------------------
1. Camera Information
--------------------------------------------------
- Resolution: 346 x 260 pixels
- Model: Pinhole (ideal projection with distortion correction)
Cameras:
- dvs (event stream)
- standard (grayscale frames)
Both share identical calibration values.
--------------------------------------------------
2. Intrinsics (fx, fy, cx, cy)
--------------------------------------------------
fx = 254.6368
fy = 253.6548
cx = 168.9970
cy = 121.0141
Meaning:
- fx, fy: focal lengths in pixel units
- cx, cy: principal point (image center)
--------------------------------------------------
3. Distortion Parameters (k1, k2, p1, p2)
--------------------------------------------------
k1 = -0.3755
k2 = 0.1311
p1 = 0.00137
p2 = -0.00159
Meaning:
- k1, k2: radial distortion (barrel/pincushion)
- p1, p2: tangential distortion (lens misalignment)
--------------------------------------------------
4. Extrinsics (T_B_C)
--------------------------------------------------
Transformation from Camera to Body (IMU) frame:
Rotation matrix (approx. identity):
[ 0.9999 -0.0122 0.0063 ]
[ 0.0122 0.9999 0.0093 ]
[ -0.0064 -0.0093 0.9999 ]
Translation vector:
x = 0.0067 m (6.7 mm)