-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathTODO.txt
28 lines (25 loc) · 1.34 KB
/
TODO.txt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
Current major issues to address:
FILTER: SIFT is performing sub-optimally (both in time and accuracy)
FILTER: SIFT has a potential divide-by-zero
FILTER: SIFT has some very very minor seem effects, possibly due ot the MATLAB compat mode
FILTER: correlate_xy (used by Edge, Frangi, and eventually SIFT) currently can't do multithreaded in-place
it is possible that pyfftw would be better than correlate_xy anyways
Still True?
TRAIN: using non-compat SIFT filters causes issues
FILTER: higher levels have much-reduced accuracy compared to MATLAB
even inten - thus does MyDownSample have issues?
Current planned optimizations:
TEST: when doing a tile-subset only copy the surrounding region into shared memory
also, only downsample that region
TEST: deadlock on leaving first parallel block when re-running CHM-test in same process
FILTER: optimize SIFT more
UTIL: optimize/multi-thread compress and/or im2double
FILTER/TEST/TRAIN: see if any algorithms would benefit from being made into CUDA vareties
primary targets: Gabor, SIFT, HOG, Frangi, LDNN-train, LDNN-test
Minor optimizations that could be done:
TEST: give each process a different number of threads when nthreads is fractional
TRAIN: save histogram?
TRAIN: subsample?
TRAIN: how should masks be downsampled?
FILTER: Haar cc_cmp_II in place?
FILTER: Edge one less intermediate?