These 48 commits are when the Protocol Buffers files have changed:
Commit: | 4a9c9b8 | |
---|---|---|
Author: | Marco Feuerstein | |
Committer: | GitHub |
Update versions, remove grpc support. (#471) * Update versions, remove grpc support. Signed-off-by: Marco Feuerstein <mfeuerstein@lyft.com> * Remove grpcio-compiler Signed-off-by: Marco Feuerstein <mfeuerstein@lyft.com> * Fix nightly Signed-off-by: Marco Feuerstein <mfeuerstein@lyft.com>
The documentation is generated from this commit.
Commit: | 94aa138 | |
---|---|---|
Author: | Marco Feuerstein | |
Committer: | GitHub |
Fix oriented beam query. (#381)
Commit: | d44c7a0 | |
---|---|---|
Author: | Marco Feuerstein | |
Committer: | GitHub |
Change back to include bounding box for s2 cells. (#365)
Commit: | 11374af | |
---|---|---|
Author: | Nikolai Morin | |
Committer: | GitHub |
Implement PointCloud trait for S2Cells (#331)
Commit: | 163aea2 | |
---|---|---|
Author: | Nikolai Morin |
Make num_points unsigned
Commit: | 1362170 | |
---|---|---|
Author: | Nikolai Morin | |
Committer: | GitHub |
S2 Meta writing prerequisites (#333)
Commit: | aca91a3 | |
---|---|---|
Author: | Nikolai Morin |
Move s2 proto functions
Commit: | 4f59e7f | |
---|---|---|
Author: | Nikolai Morin |
Output S2Meta
Commit: | f126cef | |
---|---|---|
Author: | Caterina Vitadello | |
Committer: | GitHub |
S2 Metadata (#309) * wip * meta description also for s2cells * cargo fmt * removed clippy warnings * proto naming * pr changes * pr changes * proto reserved enum 0 * wip read and cast * read to end * removing file * cargo fmt and inserting a todo * fmt * removing test *wrong pr * clippy, fmt, * proto changes * current version * fixing Node -> OctreeNode * fixed proto conversion, moved to lib * reserve proto notation for with 128 bits * WIP * update proto * removed comments * corrections/review changes * cleaning up refactoring * single variables to tuple * cargo fmt
Commit: | 5a0a4c4 | |
---|---|---|
Author: | catevita |
WIP
Commit: | bc8f8b0 | |
---|---|---|
Author: | catevita |
reserve proto notation for with 128 bits
Commit: | c50c3a4 | |
---|---|---|
Author: | catevita |
fixing Node -> OctreeNode
Commit: | 9c722be | |
---|---|---|
Author: | catevita |
merge master
Commit: | 067721c | |
---|---|---|
Author: | catevita |
proto changes
Commit: | df14dde | |
---|---|---|
Author: | catevita |
wip read and cast
Commit: | dbd7102 | |
---|---|---|
Author: | catevita |
proto reserved enum 0
Commit: | 2d74961 | |
---|---|---|
Author: | catevita |
pr changes
Commit: | 008104e | |
---|---|---|
Author: | catevita |
pr changes
Commit: | c053052 | |
---|---|---|
Author: | catevita |
proto naming
Commit: | 7124346 | |
---|---|---|
Author: | catevita |
meta description also for s2cells
Commit: | 5c8cee6 | |
---|---|---|
Author: | catevita |
wip
Commit: | 1025bba | |
---|---|---|
Author: | Nikolai Morin | |
Committer: | Nikolai Morin |
Hardcoded camera
Commit: | 7e005a3 | |
---|---|---|
Author: | Nikolai Morin | |
Committer: | Wally B. Feed |
Oriented beam query (#249) This is an implementation of oriented beam queries. An oriented beam is the same thing as an ortho frustum, so it is not as specific as it might sound. Much like the box query, it only opens nodes intersected by the queried volume to be efficient. To do this, a specialized implementation of a collision check between two oriented cubes/boxes is provided.
Commit: | da440c8 | |
---|---|---|
Author: | Marco Feuerstein | |
Committer: | Wally B. Feed |
Change to f64 projection matrix. (#247) Currently when navigating an octree covering the earth, moving around feels like going in steps. This is due to the limited f32 resolution of the translational part of the projection matrix. This PR updates the matrix to an f64 based one, removing the "stuttering".
Commit: | 256021e | |
---|---|---|
Author: | Marco Feuerstein | |
Committer: | Alexander Belyaev |
Make xrays upgradable from previous version. (#234) * Made xrays upgradable. * Shortened call.
Commit: | 06a3f9a | |
---|---|---|
Author: | Marco Feuerstein | |
Committer: | Alexander Belyaev |
Increase octree precision to double to support octrees covering the world. (#232) * WIP * Added upgrade code. * More conversion logic. * Made octree_viewer running again. * Removed unneeded folder. * Removed comment. * Removed deprecated entry. * Beautifications. * Updated version comment. * xray doesn't support old versions. * Added comment about rendering imprecision. * Improved comment.
Commit: | 335d6a0 | |
---|---|---|
Author: | Holger Rapp | |
Committer: | Wally B. Feed |
Change octree::NodeId to be a u128 (#204) This removes the limitation of octrees having a deepest level of `resolution*2**(64/3)`. This is ~2500m cube length for the default resolution of 1mm. This change modifies NodeId to be a single 128 bit unsigned integer. The highest 8 bit are the level, while the rest is used used for the index. This fits a max level of 120/3 = 40, which is ~1099512km (~86x the diameter of the earth) for the default resolution. In my tests this comes surprisingly without performance penalties on my machine (see end of PR description). The change is backwards compatible (i.e. existing octrees can be read without change), but the octree version is still bumped, since in the NodeId proto we no longer have (level, index) but instead (high_64bit, low_64bit). This also adds a new binary `upgrade_octree` to update existing octrees to the new version. Timing numbers, commands used: ~~~ target/release/build_octree /Users/hrapp/Downloads/pointcloud.ply --output_directory <dir> target/release/octree_benchmark --no-client <dir> --num-points 100000000 ~~~ Best runs out of 2 or 3, real time: | | Octree Generation | Octree benchmark | | ------------- | ------------- | ------------------ | | master | 11:57.19 | 0:06.42 | | this work | 12:10.94 | 0:05.94 | Detailed runbook in this gist: https://gist.github.com/SirVer/9ba8bd6ca02a1051d0c98ad3cef15383
Commit: | 816e15c | |
---|---|---|
Author: | Marco Feuerstein | |
Committer: | Wally B. Feed |
GRPC server with octree ids and octree factory. (#224)
Commit: | 2e5ebf8 | |
---|---|---|
Author: | Nikolai Morin | |
Committer: | Wally B. Feed |
Add all_points rpc query and benchmark for streaming points (#179) Useful for performance optimization. The GetAllPoints query is helpful because it's a common use case and just using a huge bounding box is hacky. Furthermore, in the benchmark we don't know which pointcloud, and thus bounding box, will be used.
Commit: | cf5b718 | |
---|---|---|
Author: | Avinash Kumar |
In memory reservoir sampling
Commit: | e942ea0 | |
---|---|---|
Author: | Avinash Kumar | |
Committer: | Christoph Schütte |
Rename Vector4f message to Quaternion (#165)
Commit: | 84fe965 | |
---|---|---|
Author: | Avinash Kumar | |
Committer: | Wally B. Feed |
Add comments to frustum query proto definition (#162) Add comments to frustum query proto definition
Commit: | eaafd31 | |
---|---|---|
Author: | Avinash Kumar | |
Committer: | gaschler |
Add frustum query in octree and frustum query client (#156) * Add frustum query in octree, refactor octree grpc api * Add comment for explaining the buffer size
Commit: | 995a146 | |
---|---|---|
Author: | Avinash Kumar | |
Committer: | Christoph Schütte |
Refactor octree server grpc response proto (#158)
Commit: | 704b162 | |
---|---|---|
Author: | Holger Rapp | |
Committer: | Wally B. Feed |
Adapt proto building and naming scheme to follow bazel build. (#122) Also rename grpc/ directory into the name of the crate (point_viewer_grpc).
Commit: | 792748c | |
---|---|---|
Author: | Holger Rapp | |
Committer: | Wally B. Feed |
Adaptions for bazel build. (#120)
Commit: | 2a28ce8 | |
---|---|---|
Author: | Holger Rapp | |
Committer: | Wally B. Feed |
Change gRPC octree to not send RPC for calculating visible nodes. (#115)
Commit: | 7e30419 | |
---|---|---|
Author: | thomasschiwietz | |
Committer: | Wally B. Feed |
Add GRPC function to get points in box (#102)
Commit: | f6482c0 | |
---|---|---|
Author: | thomasschiwietz | |
Committer: | Wally B. Feed |
Remove bounding cube from meta.pb (#95) This change requires to rebuild octrees.
Commit: | 10061e4 | |
---|---|---|
Author: | Holger Rapp | |
Committer: | Wally B. Feed |
Adds a binary and crate that generates xray quad trees out of an octree. (#94) Related to #90.
Commit: | ad002d7 | |
---|---|---|
Author: | thomasschiwietz | |
Committer: | Wally B. Feed |
Add Grpc function to get root bounding cube (#93) Fixes part of #86
Commit: | 21a5176 | |
---|---|---|
Author: | thomasschiwietz | |
Committer: | Wally B. Feed |
Store meta information of all nodes in root meta.pb (#85) This pull request is part of fixing #86 Things to be done in follow up pull requests: - build_octree: store NodeMeta in HashMap and remove load/store of r*.pb (meta data for each node) - proto: replace 'repeated' by 'map' - proto: get rid of bounding_box - adapt S3Octree
Commit: | fc57bf8 | |
---|---|---|
Author: | Holger Rapp | |
Committer: | Wally B. Feed |
Keep the bounding instead of the cube for an octree. (#74) This is changing the on-disk data structure, i.e. is breaking support for old octrees. They need to be regenerated. Fixes #73.
Commit: | 719717e | |
---|---|---|
Author: | Holger Rapp | |
Committer: | Wally B. Feed |
Add implementations of the Octree trait in a new gRPC crate. (#68) Related to #53.
Commit: | 4479471 | |
---|---|---|
Author: | Holger Rapp | |
Committer: | Wally B. Feed |
Small refactorings to clear up responsibility of some structs. (#66) - Make NodeId::get_stem() public. - Consistent naming: everything is named like the crate name now. - Remove stem from NodeMeta, as not every node has a on-disk representation. - Sort Cargo.toml files alphabetically.
Commit: | ec8bd8a | |
---|---|---|
Author: | Holger Rapp | |
Committer: | Wally B. Feed |
Move to using proto3. (#56) Fixes #51.
Commit: | 1d704f5 | |
---|---|---|
Author: | Holger Rapp | |
Committer: | GitHub |
Use per-node resolution to encode position data on disk. (#7)
Commit: | 35db634 | |
---|---|---|
Author: | Holger Rapp | |
Committer: | GitHub |
Use protobuf instead of JSON. (#5) I did not manage to automatically create the .rs files from the .proto at build yet, so for now they need to be manually generated.