test_kitti.yaml

# data
basedir: "./data/kitti_odometry/dataset/sequences"
sequence: 0

# Network parameters (network structure, etc. will be inherited from the training config)
training_hypers: "./treasure/hyper.json"
using_epoch: 600

# Separate tracking and meshing.
run_async: false
# Enable visualization
vis: false
resolution: 4

# These two define the range of depth observations to be cropped. Unit is meter.
depth_cut_min: 0.5
depth_cut_max: 5.0

meshing_interval: 20
integrate_interval: 5

# Mapping parameters
mapping:
  # Bound of the scene to be reconstructed
  bound_min: [-200., -800., -100.]
  bound_max: [800., 800., 200.]
  voxel_size: 4.
  # Prune observations if detected as noise.
  prune_min_vox_obs: 0
  ignore_count_th: 100
  encoder_count_th: 600.0

vis: False


# Tracking parameters
tracking:
  # An array defining how the camera pose is optimized.
  # Each element is a dictionary:
  #   For example {"n": 2, "type": [['sdf'], ['rgb', 1]]} means to optimize the summation of sdf term and rgb term
  # at the 1st level pyramid for 2 iterations.
  iter_config:
    - {"n": 10, "type": [['sdf']]}
    - {"n": 50, "type": [['sdf']]}
  sdf:
    robust_kernel: "huber"
    robust_k: 5.0
    subsample: 0.5