README.md

This directory implements a data source based on a representation of volumes and (and optional associated object surface meshes and/or skeleton representations) as static collections of files served directly over HTTP; it therefore can be used without any special serving infrastructure. In particular, it can be used with data hosted by a cloud storage provider like Google Cloud Storage or Amazon S3. Note that it is necessary, however, to either host the Neuroglancer client from the same server or enable CORS access to the data.

Each (optionally multi-scale) volume is represented as a directory tree (served over HTTP) with the following contents:

• info file in JSON format specifying the metadata.
• One subdirectory with the same name as each scale "key" value specified in the info file. Each subdirectory contains a chunked representation of the data for a single resolution.
• One subdirectory with a name equal to the "mesh" key value in the json file (only if a "mesh" key is specified, and only for segmentation volumes). This subdirectory contains metadata and triangular mesh representations of the surfaces of objects in the volume.
• One subdirectory with a name equal to the "skeletons" key value in the json file (only if a "skeletons" key is specified, and only for segmentation volumes). This subdirectory contains skeleton representations of objects in the volume.

Within neuroglancer, a precomputed data source is specified using a URL of the form: precomputed://https://host/path/to/root/directory. If the data is being served from Google Cloud Storage (GCS), precomputed://gs://bucket/path/to/root/directory may be used as an alias for precomputed://https://storage.googleapis.com/bucket/path/to/root/directory.

info JSON file specification

The root value must be a JSON object with the following members:

• "@type": If specified, must be "neuroglancer_multiscale_volume". This optional property permits automatically detecting paths to volumes, meshes, and skeletons.
• "type": One of "image" or "segmentation", specifying the type of the volume.
• "data_type": A string value equal (case-insensitively) to the name of one of the supported DataType values specified in data_type.ts. May be one of "uint8", "uint16", "uint32", "uint64", or "float32". "float32" should only be specified for "image" volumes.
• "num_channels": An integer value specifying the number of channels in the volume. Must be 1 for "segmentation" volumes.
• "scales": Array specifying information about the supported resolutions (downsampling scales) of the volume. Each element of the array is an object with the following keys:
  • "key": String value specifying the subdirectory containing the chunked representation of the volume at this scale. May also be a relative path "/"-separated path, optionally containing ".." components, which is interpreted relative to the parent directory of the "info" file.
  • "size": 3-element array [x, y, z] of integers specifying the x, y, and z dimensions of the volume in voxels.
  • "resolution": 3-element array [x, y, z] of numeric values specifying the x, y, and z dimensions of a voxel in nanometers. The x, y, and z "resolution" values must not decrease as the index into the "scales" array increases.
  • "voxel_offset": Optional. If specified, must be a 3-element array [x, y, z] of integer values specifying a translation in voxels of the origin of the data relative to the global coordinate frame. If not specified, defaults to [0, 0, 0].
  • "chunk_sizes": Array of 3-element [x, y, z] arrays of integers specifying the x, y, and z dimensions in voxels of each supported chunk size. Typically just a single chunk size will be specified as [[x, y, z]].
  • "encoding": Specifies the encoding of the chunk data. Must be a string value equal (case-insensitively) to the name of one of the supported VolumeChunkEncoding values specified in base.ts. May be one of "raw", "jpeg", or "compressed_segmentation".
  • "compressed_segmentation_block_size": This property must be specified if, and only if, "encoding" is "compressed_segmentation". If specified, it must be a 3-element [x, y, z] array of integers specifying the x, y, and z block size for the compressed segmentation encoding.
  • "sharding": If specified, indicates that volumetric chunk data is stored using the sharded format. Must be a sharding specification. If the sharded format is used, the "chunk_sizes" member must specify only a single chunk size. If unspecified, the unsharded format is used.
• "mesh": May be optionally specified if "volume_type" is "segmentation". If specified, it must be a string value specifying the name of the subdirectory containing the mesh data.
• "skeletons": May be optionally specified if "volume_type" is "segmentation". If specified, it must be a string value specifying the name of the subdirectory containing the skeleton data.

Chunked representation of volume data

For each scale and chunk size chunk_size, the volume (of voxel dimensions size = [sx, sy, sz]) is divided into a grid of grid_size = ceil(size / chunk_size) chunks.

The grid cell with grid coordinates g, where 0 <= g < grid_size, contains the encoded data for the voxel-space subvolume [begin_offset, end_offset), where begin_offset = voxel_offset + g * chunk_size and end_offset = voxel_offset + min((g + 1) * chunk_size, size). Thus, the size of each subvolume is at most chunk_size but may be truncated to fit within the dimensions of the volume. Each subvolume is conceptually a 4-dimensional [x, y, z, channel] array.

Unsharded chunk storage

If the unsharded format is used, each chunk is stored as a separate file within the path specified by the "key" property with the name "<xBegin>-<xEnd>_<yBegin>-<yEnd>_<zBegin>-<zEnd>", where:

• <xBegin>, <yBegin>, and <zBegin> are substituted with the base-10 string representations of the x, y, and z components of begin_offset, respectively; and
• <xEnd>, <yEnd>, and <zEnd> are substituted with the base-10 string representations of the x, y, and z components of end_offset, respectively.

Sharded chunk storage

If the sharded format is used, the sharded representation of the chunk data is stored within the directory specified by the "key" property. Each chunk is identified by a uint64 chunk identifier, equal to the "compressed Morton code" of the grid cell coordinates, which is used as a key to retrieve the encoded chunk data from sharded representation.

Compresed morton code

The "compressed Morton code" is a variant of the normal Morton code where bits that would be equal to 0 for all grid cells are skipped. Specifically, given the coordinates g for a grid cell, where 0 <= g < grid_size, the compressed Morton code is computed as follows:

1. Set j := 0.
2. For i from 0 to n-1, where n is the number of bits needed to encode the grid cell coordinates:
   • For dim in 0, 1, 2 (corresponding to x, y, z):
     • If 2**i <= grid_size[dim]:
       • Set output bit j of the compressed Morton code to bit i of g[dim].
       • Set j := j + 1.

Chunk encoding

The encoding of the subvolume data in each chunk file depends on the value of the "encoding" property specified for the particular scale in the info JSON file.

raw chunk encoding

The subvolume data for the chunk is stored directly in little-endian binary format in [x, y, z, channel] Fortran order (i.e. consecutive x values are contiguous) without any header. For example, if the chunk has dimensions [32, 32, 32, 1] and has "data_type": "uint32", then the chunk file should have a length of 131072 bytes.

jpeg chunk encoding

The subvolume data for the chunk is encoded as a 1- or 3-channel JPEG image. To use this encoding, the "data_type" must be "uint8" and "num_channels" must be 1 or 3. Because of the lossiness of JPEG compression, this encoding should not be used for "segmentation" volumes or "image" volumes where it is important to retain the precise values. The width and height of the JPEG image may be arbitrary, provided that the total number of pixels is equal to the product of the x, y, and z dimensions of the subvolume, and that the 1-D array obtained by concatenating the horizontal rows of the image corresponds to the flattened [x, y, z] Fortran-order representation of the subvolume.

compressed_segmentation chunk encoding

The subvolume data for the chunk is encoded using the multi-channel format compressed segmentation format. The "data_type" must be either "uint32" or "uint64". The compression block size is specified by the "compressed_segmentation_block_size" property in the info JSON file.

Mesh representation of segmented object surfaces

If the "mesh" property is specified in the info JSON file for a "segmentation" volume, then a triangular mesh representation of the surface of some or all segmented objects may be specified. Each segmented object should correspond to a set of objects with the same non-zero integer label value specified in the volume.

There are two support mesh formats: a multi-resolution mesh format in which each segmented object is represented at multiple levels of detail using a octree decomposition, and a legacy single-resolution format.

Multi-resolution mesh format

The multi-resolution object surface meshes corresponding to a segmentation are represented as a directory tree containing the following data:

• info file in JSON format specifying the metadata.
• For each segment ID for which there is a mesh representation:
  • a "manifest" file that specifies the levels of detail and octree decomposition for the object;
  • a mesh fragment data file specifying an encoded mesh representation corresponding to each octree node.

The actual storage of the manifest and mesh fragment data depends on whether the unsharded or sharded format is used.

Multi-resolution mesh info JSON file format

The info file is a JSON-format text file. The root value must be a JSON object with the following members:

• "@type": Must be "neuroglancer_multilod_draco".
• "vertex_quantization_bits": Specifies the number of bits needed to represent each vertex position coordinate within a mesh fragment. Must be 10 or 16.
• "transform": JSON array of 12 numbers specifying a 4x3 homogeneous coordinate transform from the "stored model" coordinate space to a "model" coordinate space.
• "lod_scale_multiplier": Factor by which the lod_scales values in each <segment-id>.index file are multiplied.
• "sharding": If specified, indicates that the mesh is stored using the sharded format. Must be a sharding specification. If not specified, the unsharded storage representation is used.

Multi-resolution mesh manifest file format

For each segment ID for which there is a mesh representation, there is a binary "manifest" file in the following format:

• chunk_shape: 3x float32le, specifies the x, y, and z extents of finest octree node in the "stored model" coordinate space.
• grid_origin: 3x float32le, specifies the x, y, and z origin of the octree decomposition in the "stored model" coordinate space.
• num_lods: uint32le, specifies the number of levels of detail.
• lod_scales: num_lods float32le, specifies the scale in "stored model" spatial units corresponding to each level of detail. Each scale value is multiplied by the lod_scale_multiplier value from the info JSON file.
• vertex_offsets: num_lods*3 float32le, as a C order [vertex_offsets, 3] array specifying an offset (in the "stored model" coordinate space) to add to vertex positions for each level of detail.
• num_fragments_per_lod: num_lods uint32le, specifies the number of fragments (octree nodes) for each level of detail.
• For each lod in the range [0, num_lods):
  • fragment_positions: num_fragments_per_lod[lod]*3 uint32le, C order [3, numFragments_per_lod[lod]] array specifying the x, y, and z coordinates of the octree nodes for the given lod. The node positions must be in x, y, z Z-curve order. The node corresponds to the axis-aligned bounding box within the "stored model" coordinate space with an origin of: grid_origin + [x, y, z] * chunk_shape * (2**lod) and a shape of chunk_shape * (2**lod).
  • fragment_offfsets: ``num_fragments_per_lod[lod]uint32le, specifies the size in bytes of the encoded mesh fragment in the [mesh fragment data file](#multi-resolution-mesh-fragment-data-file-format) corresponding to each octree node in thefragment_positions` array. The starting offset of the encoded mesh data corresponding to a given octree node is equal to the sum of all prior `fragment_positions` values.

Unsharded storage of multi-resolution mesh manifest

If the unsharded format is used, the manifest for each segment is stored as a separate file within the same directory as the info file under the name <segment-id>.index, where <segment-id> is the base-10 string representation of the segment ID.

Sharded storage of multi-resolution mesh manifest

If the sharded format is used, the manifest for each segment is retrieved using the segment ID as the key. The shard files are stored in the same directory as the info file.

Multi-resolution mesh fragment data file format

The mesh fragment data files consist of the concatenation of the encoded mesh data for all octree nodes specified in the manifest file, in the same order the nodes are specified in the index file, starting with lod 0. Each mesh fragment is a Draco-encoded triangular mesh with a 3-component integer vertex position attribute. Each position component j must be in the range [0, 2**vertex_quantization_bits), where a value of x corresponds to grid_origin[i] + (fragmentPosition[i] + x / (2**vertex_quantization_bits-1) * (2**lod). The built-in Draco attribute quantization is not supported.

Each mesh fragment for lod > 0 must be partitioned by a 2x2x2 grid such that no triangle crosses a grid boundary (but may be incident to a grid boundary).

Unsharded storage of multi-resolution mesh fragment data

If the unsharded format is used, the mesh mesh fragment data file is stored as a separate file within the same directory as the info file under the name <segment-id>, where <segment-id> is the base-10 string representation of the segment ID. The HTTP server must support HTTP Range requests for these files in order to allow individual fragment meshes to be retrieved.

Sharded storage of multi-resolution mesh fragment data

If the sharded format is used, the mesh fragment data file is located immediately before the manifest file in the same shard data file. The starting offset within that shard data file is not specified explicitly but may be computed from the starting offset of the manifest file and the sum of the mesh fragment sizes specified in the manifest.

Legacy single-resolution mesh format

In addition to the multi-resolution mesh format, an older single-resolution mesh format is also supported.

The surface mesh representation for a given segmented object may be split into one or more separate fragments (e.g. corresponding to subvolumes).

Within the mesh subdirectory, for each segmented object for which a surface representation is available, there is a JSON-format metadata file named <segment-id>:0, where <segment-id> is substituted with the base-10 string representation of the segment label value. This metadata file must contain an object with a "fragments" property specifying the filenames (relative to the mesh subdirectory) containing the mesh data for each fragment.

This legacy mesh format does not support a sharded storage representation.

Each fragment file is specified in the following binary format:

• The file begins with a little-endian 32-bit unsigned integer num_vertices specifying the number of vertices.
• The [x, y, z] vertex positions (as nanometer offsets within the global coordinate frame) are stored as little-endian single precision/binary32 floating point values starting at an offset of 4 bytes from the start of the file (immediately after the num_vertices value) and ending at a byte offset of 4 + 4 * 3 * num_vertices. The x, y, and z components of the vertex positions are interleaved, i.e. [x0, y0, z0, x1, y1, z1, ...].
• The number of triangles is inferred as the number of remaining bytes in the file after the vertex position data divided by 12 (the number of remaining bytes must be a multiple of 12). The triangles are specified as an array of interleaved triplets [a, b, c] of vertex indices. The vertex indices are encoded as little-endian 32-bit unsigned integers.

Skeleton representation of segmented objects

A skeleton representation of some or all segmented objects may be specified as a directory tree consisting of the following files:

• info file in JSON format specifying the metadata.
• For each segment ID for which there is a skeleton representation, a segment data file specifying the encoded skeleton for a single segment.

The actual storage of the manifest and mesh fragment data depends on whether the unsharded or sharded format is used.

Skeleton info JSON file format

The info file is a JSON-format text file. The root value must be a JSON object with the following members:

• "@type": Must be "neuroglancer_skeletons".
• "transform": JSON array of 12 numbers specifying a 4x3 homogeneous coordinate transform from the "stored model" coordinate space to a "model" coordinate space. The "stored model" coordinate space is arbitrary. The "model" coordinate space should be in nanometers. If using a "radius" attribute, the scaling applied by "transform" should be uniform.
• "vertex_attributes": JSON array specifying additional per-vertex attributes, where each array element is a JSON object with the following members:
  • "id": Attribute identifier, must be a unique, non-empty JSON string.
  • "data_type": JSON string specifying the data type, must be one of "float32", "int8", "uint8", "int16", "uint16", "int32", "uint32".
  • "num_components": JSON number specifying the number of components per vertex.
• "sharding": If specified, indicates that the mesh is stored using the sharded format. Must be a sharding specification. If not specified, the unsharded storage representation is used.

The special vertex attribute id of "radius" may be used to indicate the radius in "stored model" units; it should have a "data_type" of "float32" and "num_components" of 1.

Encoded skeleton file format

The skeleton representation for a single segment ID is a binary file with the following format:

• num_vertices: uint32le, specifies the number of vertices.
• num_edges: uint32le, specifies the number of edges.
• vertex_positions: 3*num_vertices float32le, as a C-order [num_vertices, 3] array specifying the x, y, and z vertex positions in "stored model" coordinates.
• edges: 2*num_edges uint32le, as a C-order [num_edges, 2] array specifying the source and target vertex index in the range [0, num_vertices).
• For each additional attribute in vertex_attributes:
  • attribute_data: num_vertices * num_components elements of the specified data_type in little-endian format.

Unsharded storage of encoded skeleton data

If the unsharded format is used, the encoded skeleton data is stored as a separate file within the same directory as the info file under the name <segment-id>, where <segment-id> is the base-10 segment ID.

Sharded storage of encoded skeleton data

If the sharded format is used, the encoded skeleton data is retrieved using the segment ID as the key. The shard files are stored in the same directory as the info file.

Sharded format

The unsharded multiscale volume, mesh and skeleton formats store each volumetric chunk or per-object mesh/skeleton in a separate file; in general a single file corresponds to a single unit of data that Neuroglancer may retrieve. Separate files are simple to read and write; however, if there are a large number of chunks, the resulting large number of small files can be highly inefficient with storage systems that have a high per-file overhead, as is common in many distributed storage systems. The "sharded" format avoids that problem by combining all "chunks" into a fixed number of larger "shard" files. There are several downsides to the sharded format, however:

• It requires greater complexity in the generation pipeline.
• It is not possible to re-write the data for individual chunks; the entire shard must be re-written.
• There is somewhat higher read latency due to the need to retrieve additional index information before retrieving the actual chunk data, although this latency is partially mitigated by client-side caching of the index data in Neuroglancer.

The sharded format uses a two-level index hierarchy:

• There are a fixed number of shards, and a fixed number of minishards within each shard.
• Each chunk, identified by a uint64 identifier, is mapped via a hash function to a particular shard and minishard. In the case of meshes and skeletons, the chunk identifier is simply the segment ID. In the case of volumetric data, the chunk identifier is the compressed Morton code.
• A fixed size "shard index" stored at the start of each shard file specifies for each minishard the start and end offsets within the shard file of the corresponding "minishard index".
• The variable-size "minishard index" specifies the list of chunk ids present in the minishard and the corresponding start and end offsets of the data within the shard file.

The sharded format requires that the HTTP server support HTTP Range requests.

Sharding specification

The sharding format is specified by a sharding specification in the form of a "sharding" JSON member whose value is a JSON object with the following members:

• "@type": Must be "neuroglancer_uint64_sharded_v1".
• "preshift_bits": Specifies the number of low-order bits of the chunk ID that do not contribute to the hashed chunk ID. The hashed chunk ID is computed as hash(chunk_id >> preshift_bits).
• "hash": Specifies the hash function used to map chunk IDs to shards. Must be one of:
  • "identity": The identity function.
  • "murmurhash3_x86_128": The MurmurHash3_x86_128 hash function applied to the shifted chunk ID in little endian encoding. The low 8 bytes of the resultant hash code are treated as a little endian 64-bit number.
• "minishard_bits": Specifies the number of bits of the hashed chunk ID that determine the minishard number. The number of minishards within each shard is equal to 2**minishard_bits. The minishard number is equal to bits [0, minishard_bits) of the hashed chunk id.
• "shard_bits": Specifies the number of bits of the hashed chunk ID that determine the shard number. The number of shards is equal to 2**shard_bits. The shard number is equal to bits [minishard_bits, minishard_bits+shard_bits) of the hashed chunk ID.
• "minishard_index_encoding": Specifies the encoding of the "minishard index". If specified, must be "raw" (to indicate no compression) or "gzip" (to indicate gzip compression). If not specified, equivalent to "raw".
• "data_encoding": Specifies the encoding of the actual chunk data, in the same way as "minishard_index_encoding". In the case of multiscale meshes, this encoding applies to the manifests but not to the mesh fragment data.

For each shard number in the range [0, 2**shard_bits), there is a <shard>.shard file, where <shard> is the lowercase base-16 shard number zero padded to ceil(shard_bits/4) digits.

Note that there was an earlier (obselete) version of the sharded format, which also used the same "neuroglancer_uint64_sharded_v1" identifier. The earlier format differed only in that there was a separate <shard>.index file (containing the "shard index") and a <shard>.data file (containing the remaining data) in place of the single <shard>.shard file of the current format; the <shard>.shard file is equivalent to the concatenation of the <shard>.index and <shard>.data files of the earlier version.

Shard index format

The first 2**minishard_bits * 16 bytes of each shard file is the "shard index" consisting of 2**minishard_bits entries of the form:

• start_offset: uint64le, specifies the inclusive start byte offset of the "minishard index" in the shard file.
• end_offset: uint64le, specifies the exclusive end byte offset of the "minishard index" in the shard file.

Both the start_offset and end_offset are relative to the end of the "shard index", i.e. shard_index_end = 2**minishard_bits * 16 bytes.

That is, the encoded "minishard index" for a given minishard is stored in the byte range [shard_index_end + start_offset, shard_index_end + end_offset) of the shard file. A zero-length byte range indicates that there are no chunk IDs in the minishard.

Minishard index format

The "minishard index" stored in the shard file is encoded according to the minishard_index_encoding metadata value. The decoded "minishard index" is a binary string of 24*n bytes, specifying a contiguous C-order array of [3, n] uint64le values. Values array[0, 0], ..., array[0, n-1] specify the chunk IDs in the minishard, and are delta encoded, such that array[0, 0] is equal to the ID of the first chunk, and the ID of chunk i is equal to the sum of array[0, 0], ..., array[0, i]. The size of the data for chunk i is stored as array[2, i]. Values array[1, 0], ..., array[1, n-1] specify the starting offsets in the shard file of the data corresponding to each chunk, and are also delta encoded relative to the end of the prior chunk, such that the starting offset of the first chunk is equal to shard_index_end + array[1, 0], and the starting offset of chunk i is the sum of shard_index_end + array[1, 0], ..., array[1, i] and array[2, 0], ..., array[2, i-1].

The start and size values in the minishard index specify the location in the shard file of the chunk data, which is encoded according to the data_encoding metadata value.

HTTP Content-Encoding

The normal HTTP Content-Encoding mechanism may be used by the HTTP server to transmit data in compressed form; this is particularly useful for the JSON metadata files, unsharded "raw" or "compressed_segmentation" chunk data, unsharded skeleton data, and unsharded mesh manifests, which are likely to benefit from compression and do not support other forms of compression. Some HTTP servers can perform this compression on the fly, while others, like Google Cloud Storage, require that the data be compressed ahead of time. Note that with Google Cloud Storage (and any other system that requires ahead-of-time compression), the use of Content-Encoding is not compatible with HTTP Range requests that are needed for the sharded index and data files and unsharded multi-scale mesh fragment data files; therefore, ahead-of-time compression should not be used on such files.

Example info files

{"data_type": "uint8",
 "num_channels": 1,
 "scales": [{"chunk_sizes": [[64, 64, 64]],
   "encoding": "jpeg",
   "key": "8_8_8",
   "resolution": [8, 8, 8],
   "size": [6446, 6643, 8090],
   "voxel_offset": [0, 0, 0]},
  {"chunk_sizes": [[64, 64, 64]],
   "encoding": "jpeg",
   "key": "16_16_16",
   "resolution": [16, 16, 16],
   "size": [3223, 3321, 4045],
   "voxel_offset": [0, 0, 0]},
  {"chunk_sizes": [[64, 64, 64]],
   "encoding": "jpeg",
   "key": "32_32_32",
   "resolution": [32, 32, 32],
   "size": [1611, 1660, 2022],
   "voxel_offset": [0, 0, 0]},
  {"chunk_sizes": [[64, 64, 64]],
   "encoding": "jpeg",
   "key": "64_64_64",
   "resolution": [64, 64, 64],
   "size": [805, 830, 1011],
   "voxel_offset": [0, 0, 0]},
  {"chunk_sizes": [[64, 64, 64]],
   "encoding": "jpeg",
   "key": "128_128_128",
   "resolution": [128, 128, 128],
   "size": [402, 415, 505],
   "voxel_offset": [0, 0, 0]},
  {"chunk_sizes": [[64, 64, 64]],
   "encoding": "jpeg",
   "key": "256_256_256",
   "resolution": [256, 256, 256],
   "size": [201, 207, 252],
   "voxel_offset": [0, 0, 0]},
  {"chunk_sizes": [[64, 64, 64]],
   "encoding": "jpeg",
   "key": "512_512_512",
   "resolution": [512, 512, 512],
   "size": [100, 103, 126],
   "voxel_offset": [0, 0, 0]}],
 "type": "image"}

{"data_type": "uint64",
 "mesh": "mesh",
 "num_channels": 1,
 "scales": [{"chunk_sizes": [[64, 64, 64]],
   "compressed_segmentation_block_size": [8, 8, 8],
   "encoding": "compressed_segmentation",
   "key": "8_8_8",
   "resolution": [8, 8, 8],
   "size": [6446, 6643, 8090],
   "voxel_offset": [0, 0, 0]},
  {"chunk_sizes": [[64, 64, 64]],
   "compressed_segmentation_block_size": [8, 8, 8],
   "encoding": "compressed_segmentation",
   "key": "16_16_16",
   "resolution": [16, 16, 16],
   "size": [3223, 3321, 4045],
   "voxel_offset": [0, 0, 0]},
  {"chunk_sizes": [[64, 64, 64]],
   "compressed_segmentation_block_size": [8, 8, 8],
   "encoding": "compressed_segmentation",
   "key": "32_32_32",
   "resolution": [32, 32, 32],
   "size": [1611, 1660, 2022],
   "voxel_offset": [0, 0, 0]},
  {"chunk_sizes": [[64, 64, 64]],
   "compressed_segmentation_block_size": [8, 8, 8],
   "encoding": "compressed_segmentation",
   "key": "64_64_64",
   "resolution": [64, 64, 64],
   "size": [805, 830, 1011],
   "voxel_offset": [0, 0, 0]},
  {"chunk_sizes": [[64, 64, 64]],
   "compressed_segmentation_block_size": [8, 8, 8],
   "encoding": "compressed_segmentation",
   "key": "128_128_128",
   "resolution": [128, 128, 128],
   "size": [402, 415, 505],
   "voxel_offset": [0, 0, 0]},
  {"chunk_sizes": [[64, 64, 64]],
   "compressed_segmentation_block_size": [8, 8, 8],
   "encoding": "compressed_segmentation",
   "key": "256_256_256",
   "resolution": [256, 256, 256],
   "size": [201, 207, 252],
   "voxel_offset": [0, 0, 0]},
  {"chunk_sizes": [[64, 64, 64]],
   "compressed_segmentation_block_size": [8, 8, 8],
   "encoding": "compressed_segmentation",
   "key": "512_512_512",
   "resolution": [512, 512, 512],
   "size": [100, 103, 126],
   "voxel_offset": [0, 0, 0]}],
 "type": "segmentation"}