You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I want to perform a fixed radius search with FLANN using the autotuned index (to minimize the search time). My problem involves a large number of point (about a million) at a dimension around 5 - 10 resulting to a number of nearest neighbors in the range 0-10000. I need to find (approximately) all nearest neighbors within a fixed radius. If I use the Autotuned Index, it only gives me the 5-10 nearest neighbors.
I think i found the problem: In "autotuned_index.h" the "bestSearchParams_" are used in the function "radiusSearch". When calculating the best search parameters in "estimateSearchParams" also the variable "checks" is evaluated and set to a relatively low number.
Is there an autotuned function available for the radius search that returns all nearest neighbors within the radius?
Felix
The text was updated successfully, but these errors were encountered:
Maybe I'm wrong here, but I thought that in any case, if your data has
dimension <= 10, then it's always best to use a standard kd-tree ( http://nl.mathworks.com/help/stats/createns.html), which also happens to
produce exact results.
I want to perform a fixed radius search with FLANN using the autotuned
index (to minimize the search time). My problem involves a large number of
point (about a million) at a dimension around 5 - 10 resulting to a number
of nearest neighbors in the range 0-10000. I need to find (approximately)
all nearest neighbors within a fixed radius. If I use the Autotuned Index,
it only gives me the 5-10 nearest neighbors.
I think i found the problem: In "autotuned_index.h" the
"bestSearchParams_" are used in the function "radiusSearch". When
calculating the best search parameters in "estimateSearchParams" also the
variable "checks" is evaluated and set to a relatively low number.
Is there an autotuned function available for the radius search that
returns all nearest neighbors within the radius?
Felix
—
Reply to this email directly or view it on GitHub #219.
This may be related. I just noticed that one has to use radius squared. Unfortunately, this is not mentioned in the docs. It is nevertheless consistent with FLANN also returning distances squared.
Hi,
I want to perform a fixed radius search with FLANN using the autotuned index (to minimize the search time). My problem involves a large number of point (about a million) at a dimension around 5 - 10 resulting to a number of nearest neighbors in the range 0-10000. I need to find (approximately) all nearest neighbors within a fixed radius. If I use the Autotuned Index, it only gives me the 5-10 nearest neighbors.
I think i found the problem: In "autotuned_index.h" the "bestSearchParams_" are used in the function "radiusSearch". When calculating the best search parameters in "estimateSearchParams" also the variable "checks" is evaluated and set to a relatively low number.
Is there an autotuned function available for the radius search that returns all nearest neighbors within the radius?
Felix
The text was updated successfully, but these errors were encountered: