Tasks metadata: parametrization

Aim

They aim at keeping track of the parameters used by algorithms during any ROMI task. They enable to trace back results and compare them.

Database location

Located under the metadata directory of the plant scan dataset, these JSON files contain the parameters used to run the task and are produced by each task.

Examples of task metadata JSON file names:

AnglesAndInternodes_1_0_2_0_0_1_9e87e344e6.json
Colmap_True____feature_extrac_3bbfcb1413.json
Masks_True_5_out_9adb9db801.json
TreeGraph_out__CurveSkeleton_5dca9a2821.json
Undistorted_out_____fb3e3fa0ff.json
Voxels_False___background___False_0ac9c133f7.json
Visualization.json

The tasks ids are a concatenation of the task name, the first values of the first 3 parameters sorted by parameter name and a md5hash of the name/parameters as a cananocalised json (from luigi documentation of task_id_str)

AnglesAndInternodes task

Configuration

To configure this task, we use the [AnglesAndInternodes] section in the TOML configuration file. For example:

[AnglesAndInternodes]
upstream_task = "TreeGraph"
characteristic_length = 1.0
stem_axis_inverted = false

Database location

Found under metadata/AnglesAndInternodes_*.json.

JSON example

Example of metadata/AnglesAndInternodes_*.json corresponding to the example TOML configuration file:

{
    "task_params": {
        "characteristic_length": 1.0,
        "min_elongation_ratio": 2.0,
        "min_fruit_size": 0.1,
        "number_nn": 50,
        "output_file_id": "out",
        "scan_id": "",
        "stem_axis": 2,
        "stem_axis_inverted": "False",
        "upstream_task": "ClusteredMesh"
    }
}

ClusteredMesh task

Configuration

To configure this task, we use the [Colmap] section in the TOML configuration file. For example:

[ClusteredMesh]
upstream_task = "SegmentedPointCloud"
min_vol = 1.0
min_length = 10.0

Database location

Found under metadata/ClusteredMesh_*.json.

JSON example

Example of metadata/ClusteredMesh_*.json corresponding to the example TOML configuration file:

{
    "task_params": {
        "min_length": 10.0,
        "min_vol": 1.0,
        "output_file_id": "out",
        "scan_id": "",
        "upstream_task": "SegmentedPointCloud"
    }
}

Colmap task

Configuration

To configure this task, we use the [Colmap] section in the TOML configuration file. For example:

[Colmap]
matcher = "exhaustive"
compute_dense = false

[Colmap.cli_args.feature_extractor]
"--ImageReader.single_camera" = "1"
"--SiftExtraction.use_gpu" = "1"

[Colmap.cli_args.exhaustive_matcher]
"--SiftMatching.use_gpu" = "1"

[Colmap.cli_args.model_aligner]
"--robust_alignment_max_error" = "10"

[Colmap.bounding_box] # default to None
x = [150, 650]
y = [150, 650]
z = [-90, 300]

Database location

Found under metadata/Colmap_*.json.

JSON example

Example of metadata/Colmap_*.json corresponding to the example TOML configuration file:

{
    "bounding_box": {
        "x": [
            282.3466418101626,
            590.7798175997629
        ],
        "y": [
            386.77943280470265,
            589.1307313142569
        ],
        "z": [
            22.820168903922998,
            259.7594718279537
        ]
    },
    "task_params": {
        "align_pcd": "True",
        "calibration_scan_id": "",
        "cli_args": {
            "exhaustive_matcher": {
                "--SiftMatching.use_gpu": "1"
            },
            "feature_extractor": {
                "--ImageReader.single_camera": "1",
                "--SiftExtraction.use_gpu": "1"
            },
            "model_aligner": {
                "--robust_alignment_max_error": "10"
            }
        },
        "compute_dense": "False",
        "matcher": "exhaustive",
        "output_file_id": "out",
        "scan_id": "",
        "upstream_task": "ImagesFilesetExists"
    }
}

CurveSkeleton task

Configuration

To configure this task, we use the [CurveSkeleton] section in the TOML configuration file. For example:

[CurveSkeleton]
upstream_task = "TriangleMesh"

Database location

Found under metadata/CurveSkeleton_*.json.

JSON example

Example of metadata/CurveSkeleton_*.json corresponding to the example TOML configuration file:

{
    "task_params": {
        "output_file_id": "out",
        "scan_id": "",
        "upstream_task": "TriangleMesh"
    }
}

Masks task

Configuration

To configure this task, we use the [Masks] section in the TOML configuration file. Default parametrization based on linear masking. For example:

[Masks]
upstream_task = "Undistorted"
type = "linear"
parameters = "[0,1,0]"
dilation = 5
binarize = true
threshold = 0.3

Database location

Found under metadata/Masks_*.json.

JSON example

Example of metadata/Masks_*.json corresponding to the example TOML configuration file:

{
    "task_params": {
        "binarize": "True",
        "dilation": 5,
        "output_file_id": "out",
        "parameters": [
            0,
            1,
            0
        ],
        "query": {},
        "scan_id": "",
        "threshold": 0.3,
        "type": "linear",
        "upstream_task": "ImagesFilesetExists"
    }
}

PointCloud task

Configuration

To configure this task, we use the [PointCloud] section in the TOML configuration file. For example:

[PointCloud]
upstream_task = "Voxels"
level_set_value = 1.0
log = false
background_prior= -200
min_contrast = 10.0
min_score = 0.2

Database location

Found under metadata/PointCloud_*.json.

JSON example

Example of metadata/PointCloud_*.json:

{
    "task_params": {
        "background_prior": 0.5,
        "level_set_value": 1,
        "log": "False",
        "min_contrast": 10.0,
        "min_score": 0.2,
        "output_file_id": "out",
        "scan_id": "",
        "upstream_task": "Voxels"
    }
}

Segmentation2D task

Configuration

To configure this task, we use the [Segmentation2D] section in the TOML configuration file. For example:

[Segmentation2D]
upstream_task = "Undistorted"
model_fileset = "ModelFileset"
query = "{\"channel\":\"rgb\"}"
model_id = "Resnetdataset_gl_png_896_896_epoch50"
resize = true
binarize = true
dilation = 1
Sx = 896
Sy = 896
epochs = 1
batch = 1
learning_rate = 0.0001
threshold = 0.0035

[ModelFileset]
scan_id = "models"

Database location

Found under metadata/Segmentation2D_*.json.

JSON example

Example of metadata/Segmentation2D_*.json corresponding to the example TOML configuration file:

{
    "task_params": {
        "Sx": 896,
        "Sy": 896,
        "binarize": "True",
        "dilation": 1,
        "inverted_labels": [
            "background"
        ],
        "labels": [],
        "model_fileset": "ModelFileset",
        "model_id": "Resnetdataset_gl_png_896_896_epoch50",
        "output_file_id": "out",
        "query": {
            "channel": "rgb"
        },
        "resize": "True",
        "scan_id": "",
        "threshold": 0.0035,
        "upstream_task": "Undistorted"
    }
}

SegmentedPointCloud task

Configuration

To configure this task, we use the [SegmentedPointCloud] section in the TOML configuration file. For example:

[SegmentedPointCloud]
upstream_segmentation = "Segmentation2D"
upstream_task = "PointCloud"
use_colmap_poses = true

Database location

Found under metadata/SegmentedPointCloud_*.json.

JSON example

Example of metadata/SegmentedPointCloud_*.json corresponding to the example TOML configuration file:

{
    "task_params": {
        "output_file_id": "out",
        "scan_id": "",
        "upstream_segmentation": "Segmentation2D",
        "upstream_task": "PointCloud",
        "use_colmap_poses": "True"
    }
}

TreeGraph task

Configuration

To configure this task, we use the [TreeGraph] section in the TOML configuration file. For example:

[TreeGraph]
upstream_task = "CurveSkeleton"
z_axis = 2
stem_axis_inverted = false

Database location

Found under metadata/TreeGraph_*.json.

JSON example

Example of metadata/TreeGraph_*.json corresponding to the example TOML configuration file:

{
    "task_params": {
        "output_file_id": "out",
        "scan_id": "",
        "upstream_task": "CurveSkeleton",
        "z_axis": 2,
        "z_orientation": 1
    }
}

TriangleMesh task

Configuration

To configure this task, we use the [TriangleMesh] section in the TOML configuration file. For example:

[TriangleMesh]
upstream_task = "PointCloud"

Database location

Found under metadata/TriangleMesh_*.json.

JSON example

Example of metadata/TriangleMesh_*.json corresponding to the example TOML configuration file:

{
    "task_params": {
        "output_file_id": "out",
        "scan_id": "",
        "upstream_task": "PointCloud"
    }
}

Undistorted task

Configuration

To configure this task, we use the [Undistorted] section in the TOML configuration file. For example:

[Undistorted]
upstream_task = "ImagesFilesetExists"

Database location

Found under metadata/Undistorted_*.json.

JSON example

Example of metadata/Undistorted_*.json corresponding to the example TOML configuration file:

{
    "task_params": {
        "output_file_id": "out",
        "query": {},
        "scan_id": "",
        "upstream_task": "ImagesFilesetExists"
    }
}

Visualization task

Configuration

To configure this task, we use the [Visualization] section in the TOML configuration file. For example:

[Visualization]
upstream_point_cloud = "PointCloud"
upstream_mesh = "TriangleMesh"
upstream_colmap = "Colmap"
upstream_angles = "AnglesAndInternodes"
upstream_skeleton = "CurveSkeleton"
upstream_images = "ImagesFilesetExists"
max_image_size = 1500
max_point_cloud_size = 10000
thumbnail_size = 150

Database location

Found under metadata/Visualization.json.

JSON example

Example of metadata/Visualization.json corresponding to the example TOML configuration file:

{
    "files": {
        "angles": "AnglesAndInternodes",
        "camera": "cameras",
        "images": [
            "image_00000_rgb",
            "image_00001_rgb",
            "image_00002_rgb",
            "image_00003_rgb",
            "image_00004_rgb",
            "image_00005_rgb",
            "image_00006_rgb",
            "image_00007_rgb",
            "image_00008_rgb",
            "image_00009_rgb",
            "image_00010_rgb",
            "image_00011_rgb",
            "image_00012_rgb",
            "image_00013_rgb",
            "image_00014_rgb",
            "image_00015_rgb",
            "image_00016_rgb",
            "image_00017_rgb",
            "image_00018_rgb",
            "image_00019_rgb",
            "image_00020_rgb",
            "image_00021_rgb",
            "image_00022_rgb",
            "image_00023_rgb",
            "image_00024_rgb",
            "image_00025_rgb",
            "image_00026_rgb",
            "image_00027_rgb",
            "image_00028_rgb",
            "image_00029_rgb",
            "image_00030_rgb",
            "image_00031_rgb",
            "image_00032_rgb",
            "image_00033_rgb",
            "image_00034_rgb",
            "image_00035_rgb",
            "image_00036_rgb",
            "image_00037_rgb",
            "image_00038_rgb",
            "image_00039_rgb",
            "image_00040_rgb",
            "image_00041_rgb",
            "image_00042_rgb",
            "image_00043_rgb",
            "image_00044_rgb",
            "image_00045_rgb",
            "image_00046_rgb",
            "image_00047_rgb",
            "image_00048_rgb",
            "image_00049_rgb",
            "image_00050_rgb",
            "image_00051_rgb",
            "image_00052_rgb",
            "image_00053_rgb",
            "image_00054_rgb",
            "image_00055_rgb",
            "image_00056_rgb",
            "image_00057_rgb",
            "image_00058_rgb",
            "image_00059_rgb"
        ],
        "measures": "measures",
        "mesh": "TriangleMesh",
        "point_cloud": "PointCloud",
        "poses": "images",
        "skeleton": "CurveSkeleton",
        "thumbnails": [
            "thumbnail_00000_rgb",
            "thumbnail_00001_rgb",
            "thumbnail_00002_rgb",
            "thumbnail_00003_rgb",
            "thumbnail_00004_rgb",
            "thumbnail_00005_rgb",
            "thumbnail_00006_rgb",
            "thumbnail_00007_rgb",
            "thumbnail_00008_rgb",
            "thumbnail_00009_rgb",
            "thumbnail_00010_rgb",
            "thumbnail_00011_rgb",
            "thumbnail_00012_rgb",
            "thumbnail_00013_rgb",
            "thumbnail_00014_rgb",
            "thumbnail_00015_rgb",
            "thumbnail_00016_rgb",
            "thumbnail_00017_rgb",
            "thumbnail_00018_rgb",
            "thumbnail_00019_rgb",
            "thumbnail_00020_rgb",
            "thumbnail_00021_rgb",
            "thumbnail_00022_rgb",
            "thumbnail_00023_rgb",
            "thumbnail_00024_rgb",
            "thumbnail_00025_rgb",
            "thumbnail_00026_rgb",
            "thumbnail_00027_rgb",
            "thumbnail_00028_rgb",
            "thumbnail_00029_rgb",
            "thumbnail_00030_rgb",
            "thumbnail_00031_rgb",
            "thumbnail_00032_rgb",
            "thumbnail_00033_rgb",
            "thumbnail_00034_rgb",
            "thumbnail_00035_rgb",
            "thumbnail_00036_rgb",
            "thumbnail_00037_rgb",
            "thumbnail_00038_rgb",
            "thumbnail_00039_rgb",
            "thumbnail_00040_rgb",
            "thumbnail_00041_rgb",
            "thumbnail_00042_rgb",
            "thumbnail_00043_rgb",
            "thumbnail_00044_rgb",
            "thumbnail_00045_rgb",
            "thumbnail_00046_rgb",
            "thumbnail_00047_rgb",
            "thumbnail_00048_rgb",
            "thumbnail_00049_rgb",
            "thumbnail_00050_rgb",
            "thumbnail_00051_rgb",
            "thumbnail_00052_rgb",
            "thumbnail_00053_rgb",
            "thumbnail_00054_rgb",
            "thumbnail_00055_rgb",
            "thumbnail_00056_rgb",
            "thumbnail_00057_rgb",
            "thumbnail_00058_rgb",
            "thumbnail_00059_rgb"
        ],
        "zip": "scan"
    },
    "task_params": {
        "max_image_size": 1500,
        "max_point_cloud_size": 10000,
        "output_file_id": "out",
        "scan_id": "",
        "thumbnail_size": 150,
        "upstream_angles": "AnglesAndInternodes",
        "upstream_colmap": "Colmap",
        "upstream_images": "Undistorted",
        "upstream_mesh": "TriangleMesh",
        "upstream_point_cloud": "PointCloud",
        "upstream_skeleton": "CurveSkeleton"
    }
}

Voxels task

Configuration

To configure this task, we use the [Voxels] section in the TOML configuration file. For example:

[Voxels]
upstream_mask = "Masks"
upstream_colmap = "Colmap"
use_colmap_poses = true
voxel_size = 1.0
type = "carving"
log = false
invert = false
labels = "[]"

Database location

Found under metadata/Voxels_*.json.

JSON example

Example of metadata/Voxels_*.json corresponding to the example TOML configuration file:

{
    "task_params": {
        "invert": "False",
        "labels": [
            "background"
        ],
        "log": "False",
        "output_file_id": "out",
        "scan_id": "",
        "type": "averaging",
        "upstream_colmap": "Colmap",
        "upstream_mask": "Segmentation2D",
        "use_colmap_poses": "True",
        "voxel_size": 0.01
    }
}

Scan task

Todo

For some reason the parameters are defined in the metadata/metadata.json file. Definition can be found here.