frgs 🐸 CLI Tool
fVDB-Reality-Capture provides a command-line interface (CLI) tool called frgs for tools related to Gaussian splatting.
frgs is pronounced “frogs” 🐸, and is short for Fvdb-Reality-capture Gaussian Splatting.
This tool allows users to perform Gaussian-splatting-related tasks such as reconstruction, format conversions, mesh and point cloud extraction, and visualization directly from the command line, without needing to write any code.
frgs convert
usage: frgs convert [-h] PATH PATH
Convert a Gaussian Splat in one format to another. Currently the following conversions are
supported:
- PLY to USDZ
- Checkpoint to USDZ
- PLY to PLY (copy)
- Checkpoint to PLY (export)
Example usage:
# Convert a PLY file to a USDZ file
frgs frgs convert input.ply output.usdz
# Convert a Checkpoint file to a USDZ file
frgs frgs convert input.pt output.usdz
╭─ positional arguments ─────────────────────────────────────────────────────────────────────────╮
│ PATH Path to the input file. Must be a .ply file or Checkpoint (.pt or .pth) │
│ file. (required) │
│ PATH Path to the output file. Must be a .ply file, Checkpoint (.pt or .pth) file, │
│ or .usdz file. (required) │
╰────────────────────────────────────────────────────────────────────────────────────────────────╯
╭─ options ──────────────────────────────────────────────────────────────────────────────────────╮
│ -h, --help show this help message and exit │
╰────────────────────────────────────────────────────────────────────────────────────────────────╯
frgs download
usage: frgs download [-h] [--download-path STR|PATH]
[{all,mipnerf360,gettysburg,safety_park,miris_factory}]
Download example datasets for fvdb-reality-capture. Takes in the name of a dataset and downloads
it to the specified path. If "all" is specified, all available datasets will be downloaded.
The available datasets are:
- mipnerf360: A dataset from the Mip-NeRF 360 paper.
- gettysburg: A photogrammetry dataset of the Gettysburg battlefield.
- safety_park: A photogrammetry dataset of a safety park.
- miris_factory: A dataset of synthetic images of a car factory generated by Miris.
Example usage:
# Download the mipnerf360 dataset to the ./data directory
frgs download mipnerf360 --download-path ./data
# Download all datasets to the ./datasets directory
frgs download all --download-path ./datasets
╭─ positional arguments ─────────────────────────────────────────────────────────────────────────╮
│ [{all,mipnerf360,gettysburg,safety_park,miris_factory}] │
│ The name of the dataset to download. Use "all" to download all datasets. (default: all) │
╰────────────────────────────────────────────────────────────────────────────────────────────────╯
╭─ options ──────────────────────────────────────────────────────────────────────────────────────╮
│ -h, --help │
│ show this help message and exit │
│ --download-path STR|PATH │
│ The path to download the dataset to. Data will be downloaded to a subdirectory of this │
│ path with the name of the dataset. │
│ _e.g._ `--download-path ./data` will download the Gettysburg dataset to │
│ `./data/gettysburg`. Defaults to ${CWD}/data. (default: │
│ /home/fwilliams/projects/openvdb/fvdb-reality-capture/data) │
╰────────────────────────────────────────────────────────────────────────────────────────────────╯
frgs evaluate
usage: frgs evaluate [-h] [EVALUATE OPTIONS] PATH
Evaluate a Gaussian Splat reconstruction on the validation set. You can change the validation
split used for evaluation using `--use-every-n-as-val` argument. For example,
`--use-every-n-as-val 10` will use every 10th image in the dataset as a validation image. If you
do not provide this argument, the validation split provided in the dataset (if any) will be used.
If the dataset does not provide a validation split, all images will be used for evaluation.
This will render each image in the validation set, compute statistics (PSNR, SSIM, LPIPS), and
save the rendered images and ground truth validation images to disk.
By default results will be saved to a directory named "eval" in the same directory as the
checkpoint.
Example usage:
# Evaluate a checkpoint and save results to the default log path
frgs evaluate checkpoint.pt
# Evaluate a checkpoint on a new dataset split
frgs evaluate checkpoint.pt --use-every-n-as-val 10
# Evaluate a checkpoint and save results to a custom log path
frgs evaluate checkpoint.pt --log-path ./eval_results
# Evaluate a checkpoint but don't write out rendered images
frgs evaluate checkpoint.pt --save-images False
╭─ positional arguments ─────────────────────────────────────────────────────────────────────────╮
│ PATH │
│ Path to the checkpoint file containing the Gaussian Splat model. (required) │
╰────────────────────────────────────────────────────────────────────────────────────────────────╯
╭─ options ──────────────────────────────────────────────────────────────────────────────────────╮
│ -h, --help │
│ show this help message and exit │
│ -l {None}|PATH, --log-path {None}|PATH │
│ Path to save the evaluation results. If not provided, results will be saved in a │
│ subdirectory of the checkpoint directory named "eval". (default: None) │
│ -vn {None}|INT, --use-every-n-as-val {None}|INT │
│ Use every n-th image as a validation image. If not set, will use the validation split │
│ provided in the dataset. If the dataset does not provide a validation split, will use all │
│ images for evaluation. (default: None) │
│ -s, --save-images, --no-save-images │
│ Whether to save the rendered images. Defaults to True. (default: True) │
│ --device STR|DEVICE │
│ Device to use for computation. Defaults to "cuda". (default: cuda) │
╰────────────────────────────────────────────────────────────────────────────────────────────────╯
frgs mesh-basic
usage: frgs mesh-basic [-h] [MESH-BASIC OPTIONS] PATH FLOAT
Extract a triangle mesh from a saved Gaussian splat file with TSDF fusion using depth maps
rendered from the Gaussian splat model. The algorithm proceeds in three steps:
1. First, it renders depth and color/feature images from the Gaussian splat radiance field at
each of the specified camera views.
2. Second, it integrates the depths and colors/features into a sparse fvdb.Grid in a narrow
band around the surface using sparse truncated signed distance field (TSDF) fusion. The
result is a sparse voxel grid representation of the scene where each voxel stores a signed
distance value and color (or other features).
3. Third, it extracts a mesh using the sparse marching cubes algorithm implemented in
fvdb.Grid.marching_cubes over the Grid and TSDF values. This step produces a triangle mesh
with vertex colors sampled from the colors/features stored in the Grid.
Example usage:
# Extract a mesh from a Gaussian splat model saved in `model.pt` with a truncation margin of
# 0.05
frgs mesh-basic model.pt 0.05 --output-path mesh.ply
# Extract a mesh from a Gaussian splat model saved in `model.ply` with a truncation margin of
# 0.1 with a grid shell thickness of 5 voxels, near plane at 0.1x median depth, far plane at
# 2.0x median depth of each image.
frgs mesh-basic model.ply 0.1 --output-path mesh.ply --grid-shell-thickness 5.0 --near 0.1 --far 2.0
╭─ positional arguments ─────────────────────────────────────────────────────────────────────────╮
│ PATH Path to the input PLY or checkpoint file. Must end in .ply, .pt, or │
│ .pth. (required) │
│ FLOAT Truncation margin for TSDF volume. This is the distance (in world │
│ units) that the TSDF values are truncated to. (required) │
╰────────────────────────────────────────────────────────────────────────────────────────────────╯
╭─ options ──────────────────────────────────────────────────────────────────────────────────────╮
│ -h, --help show this help message and exit │
│ -g FLOAT, --grid-shell-thickness FLOAT │
│ The number of voxels along each axis to include in the TSDF volume. │
│ This defines the resolution of the narrow band around the surface. │
│ (default: 3.0) │
│ -n FLOAT, --near FLOAT Near plane distance for which depth values are considered valid. The │
│ units depend on the `near_far_units` parameter. By default, this is a │
│ multiple of the median depth of each image. (default: 0.2) │
│ -f FLOAT, --far FLOAT Far plane distance for which depth values are considered valid. The │
│ units depend on the `near_far_units` parameter. By default, this is a │
│ multiple of the median depth of each image. (default: 1.5) │
│ -at FLOAT, --alpha-threshold FLOAT │
│ Alpha threshold to mask pixels where the Gaussian splat model is │
│ transparent, usually indicating the background. (default is 0.1). │
│ (default: 0.1) │
│ -idf INT, --image-downsample-factor INT │
│ Factor by which to downsample the rendered images for depth estimation │
│ (default is 1, _i.e._ no downsampling). (default: 1) │
│ -nfu {absolute,camera_extent,median_depth}, --near-far-units │
│ {absolute,camera_extent,median_depth} │
│ Which units to use for near and far clipping. │
│ - "absolute" means the near and far values are in world units. │
│ - "camera_extent" means the near and far values are fractions of the │
│ maximum distance from any camera to the centroid of all cameras (this │
│ is good for orbit captures). │
│ - "median_depth" means the near and far values are fractions of the │
│ median depth of each image. This is good for captures where the │
│ cameras are not evenly distributed around the scene. │
│ (default is "median_depth"). (default: median_depth) │
│ -o PATH, --output-path PATH │
│ Path to save the extracted mesh (default is "mesh.ply"). (default: │
│ mesh.ply) │
│ -d STR, --device STR Extract a mesh from a Gaussian Splat reconstruction. (default: cuda) │
╰────────────────────────────────────────────────────────────────────────────────────────────────╯
frgs mesh-dlnr
usage: frgs mesh-dlnr [-h] [MESH-DLNR OPTIONS] PATH FLOAT
Extract a mesh from a saved Gaussian splat file using TSDF fusion and depth maps estimated using
the DLNR model.
1. First, it renders stereo pairs of images from the Gaussian splat radiance field, and uses
DLNR to compute depth maps from these stereo pairs in the frame of the first image in the pair.
The result is a set of depth maps aligned with the rendered images.
2. Second, it integrates the depths and colors/features into a sparse fvdb.Grid in a narrow band
around the surface using sparse truncated signed distance field (TSDF) fusion.
The result is a sparse voxel grid representation of the scene where each voxel stores a signed
distance value and color (or other features).
3. Third, it extracts a mesh using the sparse marching cubes algorithm implemented in
fvdb.Grid.marching_cubes over the Grid and TSDF values. This step produces a triangle mesh with
vertex colors sampled from the colors/features stored in the Grid.
Example usage:
# Extract a mesh from a Gaussian splat model saved in `model.pt` with a truncation margin of
# 0.05
frgs mesh-dlnr model.pt 0.05 --output-path mesh.ply
# Extract a mesh from a Gaussian splat model saved in `model.ply` with a truncation margin of
#0.1 with a grid shell thickness of 5 voxels
frgs mesh-dlnr model.ply 0.1 --output-path mesh.ply --grid-shell-thickness 5.0
╭─ positional arguments ─────────────────────────────────────────────────────────────────────────╮
│ PATH Path to the input PLY or checkpoint file. Must end in .ply, .pt, or │
│ .pth. (required) │
│ FLOAT Truncation margin for TSDF volume. This is the distance (in world │
│ units) that the TSDF values are truncated to. (required) │
╰────────────────────────────────────────────────────────────────────────────────────────────────╯
╭─ options ──────────────────────────────────────────────────────────────────────────────────────╮
│ -h, --help show this help message and exit │
│ -g FLOAT, --grid-shell-thickness FLOAT │
│ The number of voxels along each axis to include in the TSDF volume. │
│ This defines the resolution of the narrow band around the surface. │
│ (default: 3.0) │
│ -b FLOAT, --baseline FLOAT │
│ Baseline distance (as a fraction of the mean depth of each image) used │
│ for generating stereo pairs as input to the DLNR model (default is │
│ 0.07). (default: 0.07) │
│ -n FLOAT, --near FLOAT Near plane distance (as a multiple of the baseline) for which depth │
│ values are considered valid. (default: 4.0) │
│ -f FLOAT, --far FLOAT Far plane distance (as a multiple of the baseline) for which depth │
│ values are considered valid. (default: 20.0) │
│ -at FLOAT, --alpha-threshold FLOAT │
│ Alpha threshold to mask pixels where the Gaussian splat model is │
│ transparent, usually indicating the background. (default is 0.1). │
│ (default: 0.1) │
│ -dt FLOAT, --disparity-reprojection-threshold FLOAT │
│ Reprojection error threshold for occlusion masking in pixels (default │
│ is 3.0). (default: 3.0) │
│ -idf INT, --image-downsample-factor INT │
│ Factor by which to downsample the rendered images for depth estimation │
│ (default is 1, _i.e._ no downsampling). (default: 1) │
│ -db STR, --dlnr-backbone STR │
│ Backbone to use for the DLNR model, either "middleburry" or │
│ "sceneflow" (default is "middleburry"). (default: middleburry) │
│ -ab, --use-absolute-baseline, --no-use-absolute-baseline │
│ If True, use the provided baseline as an absolute distance in world │
│ units (default is False). (default: False) │
│ -o PATH, --output-path PATH │
│ Path to save the extracted mesh (default is "mesh.ply"). (default: │
│ mesh.ply) │
│ -d STR, --device STR Extract a mesh from a Gaussian Splat reconstruction. (default: cuda) │
╰────────────────────────────────────────────────────────────────────────────────────────────────╯
frgs points
usage: frgs points [-h] [POINTS OPTIONS] PATH
Extract a point cloud with colors/features from a Gaussian splat file by unprojecting depth images
rendered from it. This algorithm can optionally filter out points near depth discontinuities using
the following heurstic:
1. Apply a small Gaussian filter to the depth images to reduce noise.
2. Run a Canny edge detector on the depth immage to find
depth discontinuities. The result is an image mask where pixels near depth edges are marked.
3. Dilate the edge mask to remove depth samples near edges.
4. Remove points from the point cloud where the corresponding depth pixel is marked in the
dilated edge mask.
Example usage:
# Extract a point cloud from a Gaussian splat model saved in `model.pt`
frgs points model.pt --output-path points.ply
# Extract a point cloud from a Gaussian splat model saved in `model.ply`
frgs points model.ply --output-path points.ply
╭─ positional arguments ─────────────────────────────────────────────────────────────────────────╮
│ PATH input-path (required) │
╰────────────────────────────────────────────────────────────────────────────────────────────────╯
╭─ options ──────────────────────────────────────────────────────────────────────────────────────╮
│ -h, --help show this help message and exit │
│ -n FLOAT, --near FLOAT Near plane distance for which depth values are considered valid. The │
│ units depend on the `near_far_units` parameter. By default, this is a │
│ multiple of the median depth of each image. (default: 0.2) │
│ -f FLOAT, --far FLOAT Far plane distance for which depth values are considered valid. The │
│ units depend on the `near_far_units` parameter. By default, this is a │
│ multiple of the median depth of each image. (default: 1.5) │
│ -at FLOAT, --alpha-threshold FLOAT │
│ Alpha threshold to mask pixels where the Gaussian splat model is │
│ transparent, usually indicating the background. (default is 0.1). │
│ (default: 0.1) │
│ -ces FLOAT, --canny-edge-std FLOAT │
│ Standard deviation for the Gaussian filter applied to the depth image │
│ before Canny edge detection (default is 1.0). Set to 0.0 to disable │
│ canny edge filtering. (default: 1.0) │
│ -cmd INT, --canny-mask-dilation INT │
│ Dilation size for the Canny edge mask (default is 5). (default: 5) │
│ -idf INT, --image-downsample-factor INT │
│ Factor by which to downsample the rendered images for depth estimation │
│ (default is 1, _i.e._ no downsampling). (default: 1) │
│ -nfu {absolute,camera_extent,median_depth}, --near-far-units │
│ {absolute,camera_extent,median_depth} │
│ Which units to use for near and far clipping. │
│ - "absolute" means the near and far values are in world units. │
│ - "camera_extent" means the near and far values are fractions of the │
│ maximum distance from any camera to the centroid of all cameras (this │
│ is good for orbit captures). │
│ - "median_depth" means the near and far values are fractions of the │
│ median depth of each image. This is good for captures where the │
│ cameras are not evenly distributed around the scene. │
│ (default is "median_depth"). (default: median_depth) │
│ -o PATH, --output-path PATH │
│ Path to save the extracted mesh (default is "mesh.ply"). (default: │
│ points.ply) │
│ -d STR, --device STR Device to use for computation (default is "cuda"). (default: cuda) │
╰────────────────────────────────────────────────────────────────────────────────────────────────╯
frgs reconstruct
usage: frgs reconstruct [-h] [RECONSTRUCT OPTIONS] PATH
Reconstruct a Gaussian Splat Radiance Field from a dataset of posed images, and save the result
as a PLY or USDZ file. Example usage:
# Reconstruct a Gaussian splat radiance field from a Colmap dataset
frgs reconstruct ./colmap_dataset -o ./output.ply
# Reconstruct a Gaussian splat radiance field from a dataset of e57 files
frgs reconstruct ./simple_directory_dataset --dataset-type e57 --out-path ./output.usdz
╭─ positional arguments ───────────────────────────────────────────────────────────────────────╮
│ PATH │
│ Path to the dataset. For "colmap" datasets, this should be the directory containing the │
│ `images` and `sparse` subdirectories. For "simple_directory" datasets, this should be │
│ the directory containing the images and a `cameras.txt` file. (required) │
╰──────────────────────────────────────────────────────────────────────────────────────────────╯
╭─ options ────────────────────────────────────────────────────────────────────────────────────╮
│ -h, --help │
│ show this help message and exit │
│ -o PATH, --out-path PATH │
│ Path to save the output PLY file. Defaults to `out.ply` in the current working │
│ directory. Path must end in .ply or .usdz. (default: out.ply) │
│ -n {None}|STR, --run-name {None}|STR │
│ Name of the run. If None, a name will be generated based on the current date and time. │
│ (default: None) │
│ -dt {colmap,simple_directory,e57}, --dataset-type {colmap,simple_directory,e57} │
│ Type of dataset to load. (default: colmap) │
│ -vn INT, --use-every-n-as-val INT │
│ Use every n-th image as a validation image. If -1, do not use a validation set. │
│ (default: -1) │
│ -uv FLOAT, --update-viz-every FLOAT │
│ How frequently (in epochs) to update the viewer during reconstruction. An epoch is one │
│ full pass through the dataset. If -1, do not visualize. (default: -1.0) │
│ -p INT, --viewer-port INT │
│ The port to expose the viewer server on if update_viz_every > 0. (default: 8080) │
│ -ip STR, --viewer-ip-address STR │
│ The IP address to expose the viewer server on if update_viz_every > 0. (default: │
│ 127.0.0.1) │
│ -d STR|DEVICE, --device STR|DEVICE │
│ Which device to use for reconstruction. Must be a cuda device. You can pass in a │
│ specific device index via cuda:N where N is the device index, or "cuda" to use the │
│ default cuda device. CPU is not supported. Default is "cuda". (default: cuda) │
│ -v, --verbose, --no-verbose │
│ If set, show verbose debug messages. (default: False) │
│ -nc INT INT INT, --nchunks INT INT INT │
│ Configuration to split the dataset into chunks for reconstruction. If set to (1, 1, 1), │
│ the dataset will not be chunked. (default: 1 1 1) │
│ -nco FLOAT, --chunk-overlap-pct FLOAT │
│ Percentage of overlap between chunks if reconstructing in chunks. Must be in [0, 1]. │
│ Only used if nchunks is not (1, 1, 1). Default is 0.1 (10% overlap). (default: 0.1) │
╰──────────────────────────────────────────────────────────────────────────────────────────────╯
╭─ cfg options ────────────────────────────────────────────────────────────────────────────────╮
│ Configuration parameters for the Gaussian splat reconstruction. │
│ ──────────────────────────────────────────────────────────────────────────────────────────── │
│ --cfg.seed INT │
│ A random seed for reproducibility. │
│ │
│ │
│ Default: 42 (the meaning of life, the universe, and everything). (default: 42) │
│ --cfg.max-epochs INT │
│ The maximum number of optimization epochs, *i.e.*, the number of times each image in the │
│ dataset will be visited. │
│ │
│ │
│ An epoch is defined as one full pass through the dataset. If you have a dataset with 100 │
│ images and a batch │
│ size of 10, then one epoch corresponds to 10 steps. │
│ │
│ │
│ Default: 200 (default: 200) │
│ --cfg.max-steps {None}|INT │
│ The maximum number of optimization steps. If set, this overrides the number of steps │
│ calculated from `max_epochs` and the dataset size. │
│ │
│ │
│ You shouldn't use this parameter unless you have a specific reason to do so. │
│ │
│ │
│ Default: None (default: None) │
│ --cfg.eval-at-percent [INT [INT ...]] │
│ Percentage of the total optimization epochs at which to perform evaluation on the │
│ validation set. │
│ │
│ │
│ For example, if `eval_at_percent` is set to `[10, 50, 100]` and `max_epochs` is set to │
│ `200`, then evaluation will be │
│ performed after 20, 100, and 200 epochs. │
│ │
│ │
│ Default: [10, 20, 30, 40, 50, 75, 100] (default: 10 20 30 40 50 75 100) │
│ --cfg.save-at-percent [INT [INT ...]] │
│ Percentage of the total optimization epochs at which to save model checkpoints. │
│ │
│ │
│ For example, if `save_at_percent` is set to `[50, 100]` and `max_epochs` is set to │
│ `200`, then checkpoints will be saved after 100 and 200 epochs. │
│ │
│ │
│ Default: [20, 100] (default: 20 100) │
│ --cfg.batch-size INT │
│ Batch size for optimization. Each step of optimization will compute losses on │
│ :obj:`batch_size` images. Note that │
│ learning rates are scaled automatically based on the batch size. │
│ │
│ │
│ Default: ``1`` (default: 1) │
│ --cfg.crops-per-image INT │
│ Number of crops to use per image during reconstruction. If you're using very large │
│ images, you can set this to a value greater than 1 │
│ to run the forward pass on crops and accumulate gradients. This can help reduce memory │
│ usage. │
│ │
│ │
│ Default: ``1`` (no cropping, use full images). (default: 1) │
│ --cfg.sh-degree INT │
│ Maximum degree of spherical harmonics to use for each Gaussian's view-dependent color. │
│ Higher degrees allow for more complex view-dependent effects, but increase memory usage │
│ and computation time. │
│ │
│ │
│ Default: ``3`` (default: 3) │
│ --cfg.increase-sh-degree-every-epoch INT │
│ When reconstructing a Gaussian splat radiance field, we start by only optimizing the │
│ diffuse (degree 0) spherical harmonics coefficients │
│ per Gaussian, and progressively increase the degree of spherical harmonics used every │
│ :obj:`increase_sh_degree_every_epoch` epochs │
│ until we reach :obj:`sh_degree`. This helps stabilize optimization in the early stages │
│ of optimization. │
│ │
│ │
│ Default: ``5`` (default: 5) │
│ --cfg.initial-opacity FLOAT │
│ Initial opacity of each Gaussian. This is the alpha value used when rendering the │
│ Gaussians at the start of optimization. │
│ │
│ │
│ Default: ``0.1`` (default: 0.1) │
│ --cfg.initial-covariance-scale FLOAT │
│ Initial scale of each Gaussian. This controls the initial size of the Gaussians in the │
│ scene. │
│ Each Gaussian's covariance matrix will be initialized to a diagonal matrix with this │
│ value on the diagonal. │
│ │
│ │
│ Default: ``1.0`` (default: 1.0) │
│ --cfg.ssim-lambda FLOAT │
│ Weight for SSIM loss. Reconstruction aims to minimize │
│ the `Structural Similarity Index Measure (SSIM) │
│ <https://en.wikipedia.org/wiki/Structural_similarity_index_measure>`_ │
│ between rendered images with the radiance field and ground truth images. This weight │
│ applies to the SSIM loss term. │
│ │
│ │
│ Default: ``0.2`` (default: 0.2) │
│ --cfg.lpips-net {vgg,alex} │
│ During evaluation, we compute the `Learned Perceptual Image Patch Similarity (LPIPS) │
│ <https://arxiv.org/abs/1801.03924>`_ metric │
│ as a measure of quality of the reconstruction. This parameter controls which network │
│ architecture is used for the LPIPS metric. │
│ │
│ │
│ Default: ``"alex"`` meaning the `AlexNet <https://en.wikipedia.org/wiki/AlexNet>`_ │
│ architecture. (default: alex) │
│ --cfg.opacity-reg FLOAT │
│ Weight for opacity regularization loss :math:`L_{opacity} = \frac{1}{N} \sum_i │
│ |opacity_i|`. │
│ │
│ │
│ If set to a value greater than 0, this will encourage the opacities of the Gaussians to │
│ be small. │
│ │
│ │
│ Default: ``0.0`` (no opacity regularization). (default: 0.0) │
│ --cfg.scale-reg FLOAT │
│ Weight for scale regularization loss :math:`L_{scale} = \frac{1}{N} \sum_i |scale_i|`. │
│ │
│ │
│ If set to a value greater than 0, this will encourage the scales of the Gaussians to be │
│ small. │
│ │
│ │
│ Default: ``0.0`` (no scale regularization). (default: 0.0) │
│ --cfg.random-bkgd, --cfg.no-random-bkgd │
│ Whether to render images with the radiance field against a background of random values │
│ during optimization. │
│ This discourages the model from using transparency to minimize loss. │
│ │
│ │
│ Default: ``False`` (default: False) │
│ --cfg.refine-start-epoch INT │
│ At which epoch to start refining the Gaussians by inserting and deleting Gaussians based │
│ on their contribution to the optimization. │
│ *e.g.* If this value is 3, the first refinement will occur at the start of epoch 3. │
│ │
│ │
│ Default: ``3`` (default: 3) │
│ --cfg.refine-stop-epoch INT │
│ At which epoch to stop refining the Gaussians by inserting and deleting Gaussians based │
│ on their contribution to the optimization. │
│ │
│ │
│ Default: ``100`` (default: 100) │
│ --cfg.refine-every-epoch FLOAT │
│ How often to refine Gaussians during optimization, in terms of epochs. │
│ For example, a value of 0.65 means refinement occurs approximately every 0.65 epochs. │
│ │
│ │
│ Default: ``0.65`` (default: 0.65) │
│ --cfg.ignore-masks, --cfg.no-ignore-masks │
│ If set to ``True``, then ignore any masks in the data and treat all pixels as valid │
│ during optimization. │
│ │
│ │
│ Default: ``False`` (default: False) │
│ --cfg.remove-gaussians-outside-scene-bbox, --cfg.no-remove-gaussians-outside-scene-bbox │
│ If set to ``True``, then Gaussians that fall outside the scene bounding box will be │
│ removed during refinement. │
│ │
│ │
│ Default: ``False`` (default: False) │
│ --cfg.optimize-camera-poses, --cfg.no-optimize-camera-poses │
│ If set to ``True``, optimize camera poses during reconstruction. This can help improve │
│ the quality of the reconstruction if the initial poses are not accurate. │
│ │
│ │
│ Default: ``True`` (default: True) │
│ --cfg.pose-opt-lr FLOAT │
│ Learning rate for camera pose optimization. │
│ │
│ │
│ Default: ``1e-5`` (default: 1e-05) │
│ --cfg.pose-opt-reg FLOAT │
│ Weight for regularization of camera pose optimization. This encourages small changes to │
│ the initial camera poses. │
│ │
│ │
│ The pose regularization loss is defined as :math:`L_{pose}` = \frac{1}{M} \sum_j │
│ ||\Delta R_j||^2 + ||\Delta t_j||^2`, │
│ *i.e.* the Frobenius norm of the change in rotation and translation for each of the │
│ ``M`` camera poses in the dataset. │
│ │
│ │
│ Default: ``1e-6`` (default: 1e-06) │
│ --cfg.pose-opt-lr-decay FLOAT │
│ Learning rate decay factor for camera pose optimization (will decay to this fraction of │
│ initial lr). │
│ │
│ │
│ Default: ``1.0`` (no decay). (default: 1.0) │
│ --cfg.pose-opt-start-epoch INT │
│ At which epoch to start optimizing camera poses. │
│ │
│ │
│ Default: ``0`` (start from beginning of optimization). (default: 0) │
│ --cfg.pose-opt-stop-epoch INT │
│ At which epoch to stop optimizing camera poses. │
│ │
│ │
│ Default: ``max_epochs`` (optimize poses for the entire duration of optimization). │
│ (default: 200) │
│ --cfg.pose-opt-init-std FLOAT │
│ Standard deviation for the normal distribution used to initialize the embeddings for │
│ camera pose optimization. │
│ │
│ │
│ Default: ``1e-4`` (default: 0.0001) │
│ --cfg.near-plane FLOAT │
│ Near plane clipping distance when rendering the Gaussians. │
│ │
│ │
│ Default: ``0.01`` (default: 0.01) │
│ --cfg.far-plane FLOAT │
│ Far plane clipping distance when rendering the Gaussians. │
│ │
│ │
│ Default: ``1e10`` (default: 10000000000.0) │
│ --cfg.min-radius-2d FLOAT │
│ Minimum screen space radius (in pixels) below which Gaussians are ignored after │
│ projection. │
│ │
│ │
│ Default: ``0.0`` (default: 0.0) │
│ --cfg.eps-2d FLOAT │
│ Amount of padding (in pixels) to add to the screen space bounding box of each Gaussian │
│ when determining which pixels it affects. │
│ │
│ │
│ Default: ``0.3`` (default: 0.3) │
│ --cfg.antialias, --cfg.no-antialias │
│ Whether to use anti-aliasing when rendering the Gaussians. │
│ │
│ │
│ Default: ``False`` (default: False) │
│ --cfg.tile-size INT │
│ Tile size (in pixels) to use when rendering the Gaussians. │
│ You should generally leave this at the default value unless you have a specific reason │
│ to change it. │
│ │
│ │
│ Default: ``16`` (default: 16) │
╰──────────────────────────────────────────────────────────────────────────────────────────────╯
╭─ tx options ─────────────────────────────────────────────────────────────────────────────────╮
│ Configuration for the transforms to apply to the scene before reconstruction. │
│ ──────────────────────────────────────────────────────────────────────────────────────────── │
│ --tx.image-downsample-factor INT │
│ Downsample images by this factor (default: 4) │
│ --tx.rescale-jpeg-quality INT │
│ JPEG quality to use when resaving images after downsampling (default: 95) │
│ --tx.points-percentile-filter FLOAT │
│ Percentile of points to filter out based on their distance from the median point │
│ (default: 0.0) │
│ --tx.normalization-type {none,pca,ecef2enu,similarity} │
│ Type of normalization to apply to the scene (default: pca) │
│ --tx.crop-bbox {None}|{FLOAT FLOAT FLOAT FLOAT FLOAT FLOAT} │
│ Optional bounding box (in the normalized space) to crop the scene to (xmin, xmax, ymin, │
│ ymax, zmin, zmax) (default: None) │
│ --tx.crop-to-points, --tx.no-crop-to-points │
│ Whether to crop the scene to the bounding box or not (default: False) │
│ --tx.min-points-per-image INT │
│ Minimum number of 3D points that must be visible in an image for it to be included in │
│ the optimization (default: 5) │
╰──────────────────────────────────────────────────────────────────────────────────────────────╯
╭─ opt options ────────────────────────────────────────────────────────────────────────────────╮
│ Configuration for the optimizer used to reconstruct the Gaussian splat radiance field. │
│ ──────────────────────────────────────────────────────────────────────────────────────────── │
│ --opt.max-gaussians INT │
│ The maximum number of Gaussians to allow in the model. If -1, no limit. (default: -1) │
│ --opt.insertion-grad-2d-threshold-mode │
│ {CONSTANT,PERCENTILE_FIRST_ITERATION,PERCENTILE_EVERY_ITERATION} │
│ Whether to use a fixed threshold for :obj:`insertion_grad_2d_threshold` (constant), a │
│ value computed as a percentile of │
│ the distribution of screen space mean gradients on the first iteration, or a percentile │
│ value │
│ computed at each refinement step. │
│ │
│ │
│ See :class:`InsertionGrad2dThresholdMode` for details on the available modes. (default: │
│ CONSTANT) │
│ --opt.deletion-opacity-threshold FLOAT │
│ If a Gaussian's opacity drops below this value, delete it during refinement. (default: │
│ 0.005) │
│ --opt.deletion-scale-3d-threshold FLOAT │
│ If a Gaussian's 3d scale is above this value, then delete it during refinement. │
│ (default: 0.1) │
│ --opt.deletion-scale-2d-threshold FLOAT │
│ If the maximum projected size of a Gaussian between refinement steps exceeds this value │
│ then delete it during │
│ refinement. │
│ │
│ │
│ .. note:: This parameter is only used if set │
│ :obj:`use_screen_space_scales_for_refinement_until` is greater than 0. (default: 0.15) │
│ --opt.insertion-grad-2d-threshold FLOAT │
│ Threshold value on the accumulated norm of projected mean gradients between refinement │
│ steps to │
│ determine whether a Gaussian has high error and is a candidate for duplication or │
│ splitting. │
│ │
│ │
│ .. note:: If :obj:`insertion_grad_2d_threshold_mode` is │
│ :obj:`InsertionGrad2dThresholdMode.CONSTANT`, then this value │
│ is used directly as the threshold, and **must be positive**. │
│ │
│ │
│ .. note:: If :obj:`insertion_grad_2d_threshold_mode` is │
│ :obj:`InsertionGrad2dThresholdMode.PERCENTILE_FIRST_ITERATION` │
│ or :obj:`InsertionGrad2dThresholdMode.PERCENTILE_EVERY_ITERATION`, then this │
│ value must be in the │
│ range ``(0.0, 1.0)`` (exclusive). (default: 0.0002) │
│ --opt.insertion-scale-3d-threshold FLOAT │
│ Duplicate high-error (determined by :obj:`insertion_grad_2d_threshold`) Gaussians whose │
│ 3d scale is below this value. │
│ These Gaussians are too small to capture the detail in the region they cover, so we │
│ duplicate them to │
│ allow them to specialize. (default: 0.01) │
│ --opt.insertion-scale-2d-threshold FLOAT │
│ Split high-error (determined by :obj:`insertion_grad_2d_threshold`) Gaussians whose │
│ maximum projected │
│ size exceeds this value. These Gaussians are too large to capture the detail in the │
│ region they cover, │
│ so we split them to allow them to specialize. │
│ │
│ │
│ .. note:: This parameter is only used if set │
│ :obj:`use_screen_space_scales_for_refinement_until` is greater than 0. (default: 0.05) │
│ --opt.opacity-updates-use-revised-formulation, │
│ --opt.no-opacity-updates-use-revised-formulation │
│ When splitting Gaussians, whether to update the opacities of the new Gaussians using the │
│ revised formulation from │
│ `*"Revising Densification in Gaussian Splatting"* <https://arxiv.org/abs/2404.06109>`_. │
│ This removes a bias which weighs newly split Gaussians contribution to the image more │
│ heavily than │
│ older Gaussians. (default: False) │
│ --opt.insertion-split-factor INT │
│ When splitting Gaussians during insertion, this value specifies the total number of new │
│ Gaussians that will │
│ replace each selected source Gaussian. The original is removed and replaced by │
│ :obj:`insertion_split_factor` new │
│ Gaussians. *e.g.* if this value is 2, each split Gaussian is replaced by 2 new smaller │
│ Gaussians │
│ (the original is removed). This value must be >= 2. (default: 2) │
│ --opt.insertion-duplication-factor INT │
│ When duplicating Gaussians during insertion, this value specifies the total number of │
│ copies (including │
│ the original) that will result for each selected source Gaussian. The original is kept, │
│ and │
│ ``insertion_duplication_factor - 1`` new identical copies are added. *e.g.* if this │
│ value is 3, │
│ each duplicated Gaussian becomes 3 copies of itself (the original plus 2 new). This │
│ value must be >= 2. (default: 2) │
│ --opt.reset-opacities-every-n-refinements INT │
│ If set to a positive value, then clamp all opacities to be at most twice the value of │
│ :obj:`deletion_opacity_threshold` every time :func:`GaussianSplatOptimizer.refine` is │
│ called │
│ :obj:`reset_opacities_every_n_refinements` times. This prevents Gaussians from becoming │
│ completely occluded by │
│ denser Gaussians and thus unable to be optimized. (default: 30) │
│ --opt.use-scales-for-deletion-after-n-refinements INT │
│ If set to a positive value, then after ``use_scales_for_deletion_after_n_refinements`` │
│ calls to │
│ :func:`GaussianSplatOptimizer.refine`, use the 3D scales of the Gaussians to determine │
│ whether to delete them. │
│ This will delete Gaussians that have grown │
│ too large in 3D space and are not contributing to the optimization. │
│ │
│ │
│ By default, this value matches :obj:`reset_opacities_every_n_refinements` so that both │
│ behaviors are enabled at the │
│ same time. (default: 30) │
│ --opt.use-screen-space-scales-for-refinement-until INT │
│ If set to a positive value, then use threshold the maximum projected size of Gaussians │
│ between refinement steps │
│ to decide whether to split or delete Gaussians that are too large. This behavior is │
│ enabled until │
│ :func:`GaussianSplatOptimizer.refine` has been called │
│ ``use_screen_space_scales_for_refinement_until`` times. │
│ After that, only 3D scales are used for refinement. (default: 0) │
│ --opt.spatial-scale-mode │
│ {ABSOLUTE_UNITS,MEDIAN_CAMERA_DEPTH,MAX_CAMERA_DEPTH,MAX_CAMERA_TO_CENTROID,SCENE_DIAGONAL_P │
│ ERCENTILE} │
│ How to interpret 3D optimization scale thresholds and learning rates (*i.e.* │
│ :obj:`insertion_scale_3d_threshold`, │
│ :obj:`deletion_scale_3d_threshold`, and :obj:`means_lr`). These are scaled by a spatial │
│ scale computed from │
│ the scene, so they are relative to the size of the scene being optimized. │
│ │
│ │
│ See :class:`SpatialScaleMode` for details on the available modes. (default: │
│ MEDIAN_CAMERA_DEPTH) │
│ --opt.spatial-scale-multiplier FLOAT │
│ Multiplier to apply to the spatial scale computed from the scene to get a slightly │
│ larger scale. (default: 1.1) │
│ --opt.means-lr FLOAT │
│ Learning rate for the means of the Gaussians. This is also scaled by the spatial scale │
│ computed from the scene. │
│ │
│ │
│ See :obj:`spatial_scale_mode` for details on how the spatial scale is computed. │
│ (default: 0.00016) │
│ --opt.log-scales-lr FLOAT │
│ Learning rate for the log scales of the Gaussians. (default: 0.005) │
│ --opt.quats-lr FLOAT │
│ Learning rate for the quaternions of the Gaussians. (default: 0.001) │
│ --opt.logit-opacities-lr FLOAT │
│ Learning rate for the logit opacities of the Gaussians. (default: 0.05) │
│ --opt.sh0-lr FLOAT │
│ Learning rate for the diffuse spherical harmonics (order 0). (default: 0.0025) │
│ --opt.shN-lr FLOAT │
│ Learning rate for the specular spherical harmonics (order > 0). (default: 0.000125) │
╰──────────────────────────────────────────────────────────────────────────────────────────────╯
╭─ io options ─────────────────────────────────────────────────────────────────────────────────╮
│ Configure saving and logging metrics, images, and checkpoints. │
│ ──────────────────────────────────────────────────────────────────────────────────────────── │
│ --io.save-images, --io.no-save-images │
│ Whether to save images to disk. If ``False``, images will not be saved to disk. │
│ │
│ │
│ Default is ``False``. (default: False) │
│ --io.save-checkpoints, --io.no-save-checkpoints │
│ Whether to save checkpoints to disk. If ``False``, checkpoints will not be saved to │
│ disk. │
│ │
│ │
│ Default is ``True``. (default: True) │
│ --io.save-plys, --io.no-save-plys │
│ Whether to save PLY files to disk. If ``False``, PLY files will not be saved to disk. │
│ │
│ │
│ Default is ``True``. (default: True) │
│ --io.save-metrics, --io.no-save-metrics │
│ Whether to save metrics to a CSV file. If ``False``, metrics will not be saved to a CSV │
│ file. │
│ │
│ │
│ Default is ``True``. (default: True) │
│ --io.metrics-file-buffer-size INT │
│ How much buffering (in bytes) to use for metrics file logging. Larger values can improve │
│ performance when logging many metrics. │
│ │
│ │
│ Default is 8 MiB. (default: 8388608) │
│ --io.use-tensorboard, --io.no-use-tensorboard │
│ Whether to use TensorBoard for logging metrics and images. If ``True``, metrics and │
│ images will be logged to TensorBoard. │
│ │
│ │
│ Default is ``False``. (default: False) │
│ --io.save-images-to-tensorboard, --io.no-save-images-to-tensorboard │
│ Whether to also save images to TensorBoard if :obj:`use_tensorboard` is ``True``. If │
│ ``True``, images will be saved to TensorBoard. │
│ │
│ │
│ Default is ``False``. (default: False) │
│ --io.log-path {None}|PATH │
│ Path to save logs, checkpoints, and other output to. Defaults to `frgs_logs` in the │
│ current working directory. (default: frgs_logs) │
│ --io.log-every INT │
│ How frequently to log metrics during reconstruction. (default: 10) │
╰──────────────────────────────────────────────────────────────────────────────────────────────╯
frgs resume
usage: frgs resume [-h] [RESUME OPTIONS] PATH
Resume reconstructing a 3D Gaussian Splat radiance field from a checkpoint. This command loads a
model checkpoint and continues reconstruction from that point. The dataset used to create the
checkpoint must be at the same path as when the checkpoint was created.
Example usage:
# Resume reconstruction from a checkpoint and save the final model to out_resumed.ply
frgs resume checkpoint.pt -o out_resumed.ply
╭─ positional arguments ───────────────────────────────────────────────────────────────────────╮
│ PATH │
│ Path to the checkpoint file containing the Gaussian Splat radiance field. (required) │
╰──────────────────────────────────────────────────────────────────────────────────────────────╯
╭─ options ────────────────────────────────────────────────────────────────────────────────────╮
│ -h, --help │
│ show this help message and exit │
│ -n {None}|STR, --run-name {None}|STR │
│ Name of the run. If None, a name will be generated based on the current date and time. │
│ (default: None) │
│ -uv FLOAT, --update-viz-every FLOAT │
│ How frequently (in epochs) to update the viewer during reconstruction. An epoch is one │
│ full pass through the dataset. If -1, do not visualize. (default: -1.0) │
│ -p INT, --viewer-port INT │
│ The port to expose the viewer server on if update_viz_every > 0. (default: 8080) │
│ -ip STR, --viewer-ip-address STR │
│ The IP address to expose the viewer server on if update_viz_every > 0. (default: │
│ 127.0.0.1) │
│ -d STR|DEVICE, --device STR|DEVICE │
│ Which device to use for reconstruction. Must be a cuda device. You can pass in a │
│ specific device index via cuda:N where N is the device index, or "cuda" to use the │
│ default cuda device. CPU is not supported. Default is "cuda". (default: cuda) │
│ -v, --verbose, --no-verbose │
│ If set, show verbose debug messages. (default: False) │
│ -o PATH, --out-path PATH │
│ Path to save the output PLY file. Defaults to `out.ply` in the current working │
│ directory. Path must end in .ply or .usdz. (default: out_resumed.ply) │
╰──────────────────────────────────────────────────────────────────────────────────────────────╯
╭─ io options ─────────────────────────────────────────────────────────────────────────────────╮
│ Configure saving and logging metrics, images, and checkpoints. │
│ ──────────────────────────────────────────────────────────────────────────────────────────── │
│ --io.save-images, --io.no-save-images │
│ Whether to save images to disk. If ``False``, images will not be saved to disk. │
│ │
│ │
│ Default is ``False``. (default: False) │
│ --io.save-checkpoints, --io.no-save-checkpoints │
│ Whether to save checkpoints to disk. If ``False``, checkpoints will not be saved to │
│ disk. │
│ │
│ │
│ Default is ``True``. (default: True) │
│ --io.save-plys, --io.no-save-plys │
│ Whether to save PLY files to disk. If ``False``, PLY files will not be saved to disk. │
│ │
│ │
│ Default is ``True``. (default: True) │
│ --io.save-metrics, --io.no-save-metrics │
│ Whether to save metrics to a CSV file. If ``False``, metrics will not be saved to a CSV │
│ file. │
│ │
│ │
│ Default is ``True``. (default: True) │
│ --io.metrics-file-buffer-size INT │
│ How much buffering (in bytes) to use for metrics file logging. Larger values can improve │
│ performance when logging many metrics. │
│ │
│ │
│ Default is 8 MiB. (default: 8388608) │
│ --io.use-tensorboard, --io.no-use-tensorboard │
│ Whether to use TensorBoard for logging metrics and images. If ``True``, metrics and │
│ images will be logged to TensorBoard. │
│ │
│ │
│ Default is ``False``. (default: False) │
│ --io.save-images-to-tensorboard, --io.no-save-images-to-tensorboard │
│ Whether to also save images to TensorBoard if :obj:`use_tensorboard` is ``True``. If │
│ ``True``, images will be saved to TensorBoard. │
│ │
│ │
│ Default is ``False``. (default: False) │
│ --io.log-path {None}|PATH │
│ Path to save logs, checkpoints, and other output to. Defaults to `frgs_logs` in the │
│ current working directory. (default: frgs_logs) │
│ --io.log-every INT │
│ How frequently to log metrics during reconstruction. (default: 10) │
╰──────────────────────────────────────────────────────────────────────────────────────────────╯
frgs show-data
usage: frgs show-data [-h] [SHOW-DATA OPTIONS] PATH
Visualize a scene in a dataset folder. This will plot the scene's cameras and point cloud in an
interactive viewer shown in a browser window.
The dataset folder should either contain a colmap dataset, a set of e57 files, a
simple_directory dataset:
COLMAP Data format: A folder should containing:
- cameras.txt
- images.txt
- points3D.txt
- A folder named "images" containing the image files.
- An optional "masks" folder with the same layout as images containing masks of which pixels
are valid.
E57 format: A folder containing one or more .e57 files.
Simple Directory format: A folder containing:
- images/ A directory of images (jpg, png, etc).
- An optional "masks/" folder with the same layout as images containing masks of which
pixels are valid.
- A cameras.json file containing camera intrinsics and extrinsics for each image. It should
be a list of objects with the following format:
"camera_name": "camera_0000",
"width": 2048,
"height": 2048,
"camera_intrinsics": [], # 3x3 matrix in row-major order
"world_to_camera": [], # 4x4 matrix in row-major order
"image_path": "name_of_image_file_relative_to_images_folder"
Example usage:
# Visualize a Colmap dataset in the folder ./colmap_dataset
frgs show-data ./colmap_dataset
# Visualize an e57 dataset in the folder ./e57_dataset
frgs show-data ./e57_dataset --dataset-type e57
# Visualize a simple directory dataset in the folder ./simple_directory_dataset
frgs show-data ./simple_directory_dataset --dataset-type simple_directory
# Flip the up axis of the scene from -Z to +Z
# It's -fu because that's what you say when your scene is backwards.
frgs show-data ./colmap_dataset -fu
╭─ positional arguments ───────────────────────────────────────────────────────────────────────╮
│ PATH │
│ Path to the dataset folder. (required) │
╰──────────────────────────────────────────────────────────────────────────────────────────────╯
╭─ options ────────────────────────────────────────────────────────────────────────────────────╮
│ -h, --help │
│ show this help message and exit │
│ -p INT, --viewer-port INT │
│ The port to expose the viewer server on. (default: 8080) │
│ -ip STR, --viewer-ip-address STR │
│ The port to expose the viewer server on. (default: 127.0.0.1) │
│ -v, --verbose, --no-verbose │
│ If True, then the viewer will log verbosely. (default: False) │
│ -ppf FLOAT, --points-percentile-filter FLOAT │
│ Percentile filter for points. Points with any coordinate below this percentile or above │
│ (100 - this percentile) will be removed from the point cloud. This can help remove │
│ outliers. Set to 0.0 to disable. (default: 0.0) │
│ -mpi INT, --min-points-per-image INT │
│ Minimum number of points a camera must observe to be included in the viewer. (default: │
│ 5) │
│ -dt {colmap,simple_directory,e57}, --dataset-type {colmap,simple_directory,e57} │
│ Type of dataset to load. (default: colmap) │
│ -al FLOAT, --axis-length FLOAT │
│ The length (in world units) of the axes drawn at each camera and at the origin. │
│ (default: 1.0) │
│ -fl FLOAT, --frustum-length FLOAT │
│ Frustum length from the origin to the view plane in world units. (default: 1.0) │
│ -flw FLOAT, --frustum-line-width FLOAT │
│ Scren space line width of the camera frustums. (default: 2.0) │
│ -alw FLOAT, --axis-line-width FLOAT │
│ Screen space line width of the axes. (default: 1.0) │
│ -fu, --flip-up-axis, --no-flip-up-axis │
│ If true, flip the up axis of the scene from -Z to +Z (default: False) │
│ -ps FLOAT, --point-size FLOAT │
│ Size of the points in screen space. (default: 1.0) │
│ -pc {None}|{FLOAT FLOAT FLOAT}, --points-color {None}|{FLOAT FLOAT FLOAT} │
│ If set to a color tuple, use this color for all points instead of their RGB values. │
│ Color values must be in the range [0.0, 1.0]. (default: None) │
╰──────────────────────────────────────────────────────────────────────────────────────────────╯
frgs show
usage: frgs show [-h] [SHOW OPTIONS] PATH
Visualize a Gaussian splat radiance field in a saved PLY or checkpoint file. This will plot the
splats in an interactive viewer shown in a browser window.
# Example usage:
# Visualize a Gaussian splat model saved in `model.ply`
frgs show model.ply --viewer-port 8888
# Visualize a Gaussian splat model saved in `model.pt`
frgs show model.pt --viewer-port 8888
╭─ positional arguments ───────────────────────────────────────────────────────────────────────╮
│ PATH Path to the input PLY or checkpoint file. Must end in .ply, .pt, or │
│ .pth. (required) │
╰──────────────────────────────────────────────────────────────────────────────────────────────╯
╭─ options ────────────────────────────────────────────────────────────────────────────────────╮
│ -h, --help show this help message and exit │
│ -p INT, --viewer-port INT │
│ The port to expose the viewer server on. (default: 8080) │
│ -ip STR, --viewer-ip-address STR │
│ The port to expose the viewer server on. (default: 127.0.0.1) │
│ -v, --verbose, --no-verbose │
│ If True, then the viewer will log verbosely. (default: False) │
│ --device STR|DEVICE Device to use for computation (default is "cuda"). (default: cuda) │
╰──────────────────────────────────────────────────────────────────────────────────────────────╯