| Age | Commit message (Collapse) | Author |
|---|
|
* unstable:
main: use camera-based centering by default
settings: clarify center method usage
accela: fix typo
win32: try win_key with modifier first
shortcuts: remove obsolete code on win32
shortcuts: allow for numpad on win32
qfc: drop nan check
pt: drop nan check
shortcuts: allow for numlock on win32
accela: remove too many nan checks
timer: guard against overflow on win32
accela: remove "done" logic
timer: sprinkle some const
shortcuts: allow for binding same key to multiple functions
qfc: guard against unlikely division by zero
cmake: update toolchain file
|
|
Bug cause was different
|
|
* unstable:
cmake: update toolchain file
shortcuts: fix osx/linux keystrokes persisting
cmake: fix copy-paste comment
cmake: add toolchain file for OSX
cmake: add osx policy to make it shutup
x-plane: ignore diagnostic
osx: nix warning
gitattributes: more text extensions to eol=lf
cmake: timestamp logic simplify/fix
cmake: no timestamp for tag builds
all: update copyright where appropriate
all: comments only
cmake: regen before making tarball
cmake: fix dropbox share invocation
tracker: initialize newpose
pt: use previous pose on NaN result from POSIT
accela: also don't poison ewma state with nans
cmake: fix tarball invocation
accela: elide NaN output values
qfc: elide NaN values
pt: reformat more
pt: reformat posit
pt: refactor auto threshold somewhat
pt: rename ill-chosen name
pt: switch min/max point size to reals
cmake: upload tarball to Dropbox but only if I'm the user
cmake: add tarball timestamp so it gets rebuilt
cmake: don't regen version if none changed
cmake: fix git describe --dirty
cmake: retab git module
fix tarball target
cmake: mark dirty tree
cmake: regen tarball even if exists
cmake: generate version.cc
tracker: check for NaN values
accela: don't check NaNs in filter, wrong place
cmake: drop -ffast-math, allow for NaN check
shortcuts: actually print screen binding works
shortcuts: alias right modifier keys to left modifier keys
shortcuts: allow for binding scroll lock and pause/break
accela: also filter out NaNs on tracking start
allow for filter immediate center
|
|
|
|
|
|
This will reset dynamic pose resolution if the error persists.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Sponsored-by: TrackHat
|
|
* unstable:
pt: add histogram-based thresholding
|
|
Otsu-based thresholding doesn't work in normal lighting
conditions.
Sponsored-by: TrackHat
|
|
* unstable:
pt: experimental auto threshold support
|
|
Sponsored-by: TrackHat
|
|
Sponsored-by: TrackHat
|
|
Sponsored-by: TrackHat
* unstable: (112 commits)
Revert "accela: try more complex deadzone algorithm"
pt: add wiki link for calibration instructions
qfc: more gray colors when graph is disabled
qfc, mapping: disallow editing and change color when checkbox disabled
qfc: don't allow editing when disabled
plugin-support: workaround multiple copies of modules
accela: change translation gain
accela: change rotation gain
pt: change extraction code
freepie-udp: add license
freepie-udp: update
accela: try more complex deadzone algorithm
accela: add helpful comment
accela: revert to more quadratic spline at start
also save if save pending when start tracking
ui: force saving if saving timer is pending
accela: slightly more linear function
pose-widget: simplify
freepie-udp: nix harmless warning
pose-widget: add comment
simple-mat: add __restrict for GNU
ui: save settings on a timer
ui: show a warning if configuration directory can't be used
aruco: tune min/max marker size
win32-joystick: fix duplicate device name selection logic
ui: no need to reset settings on tracking start
ui: save profile when combobox text changes, not merely index
ui: save current tracker/filter/proto, not last
win32-joystick: fix COM failure path
win32-joystick: nix unused variable
win32-joystick: allow for unique selection of joysticks
qfc: also draw line from function start while moving
no need to save on tracking start anymore
save axis opts, not just spline control points
tracker: change centering order again
fix typo
cmake: don't say "opentrack" twice for osx build
ui: really don't switch profiles while refreshing combobox
pt: fix type mismatch
ui: save prior to switching configs
tracker_rs: adjusted tracker's name
ui: don't reload settings if config wasn't changed
ui: save everything before switching profiles
options: return .ini basename easier
tracker_rs: more readable icon
tracker_rs: prevent potential thread starvation
ui: switch to newly-created config
ui: decruft
ui: also bail on new config if its name is ".ini"
ui: auto-refresh config list
ui: display up to 20 profiles w/o scrolling
ui: create an empty config properly in another code branch
udp-tracker: silence harmless warning
plugin-support: fix typo
plugin-support: set more sane RTLD flags
plugin-support: free modules from address space when needed
ui: rename symbol to reflect current usage
ui: move some slots to private
ui, shortcuts: these are slots, actually
ui: no need to write anything to new empty config
ui: avoid potential memory leak
ft: new game support
game-data: chomp newlines in extractor
ui: append extension in the right place
simple-mat: fix arglist SFINAE
simple-mat: can use static_cast here
nix default empty config. it gets created on its own.
ui: these aren't qt slots
ui: nix wrong headers
ui: remove GNU GPL
ui: simplify condition
props changed only
main, ui: store config files in a predefined directory
pt: show extracted points' areas and success count
close dialogs on switch to new module
...
|
|
|
|
Point size is now specified by its radius.
Points are extracted in one pass, using multiple passes caused missed
detection for some users.
Circularity is now displayed as part of overlay.
Points are now sorted by circularity, which is our confidence metric.
Tracker code ignores additional points, keeping only those with the
highest confidence metric.
|
|
|
|
Some source files had executable bit on. Remove it.
|
|
|
|
Aspect ratio is meant to be viewport's ratio, not diagonal to axis.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
This reverts commit e89c2b1034692eb1365715a089b7035e6d837719.
Issue: #199
|
|
|
|
PT tracker returns garbage when calibrating, Z is zero. Maybe can be
reverted and fixed later.
|
|
|
|
We're the only maintainer of the PT tracker now, so don't clutter the
file so much.
|
|
|
|
We can crash if start/stop pressed fast eventually.
|
|
|
|
|
|
|
|
|
|
Issue: #190
|
|
Store points as pixel coordinates, not -0.5->0.5. This has no effect on
tracking as POSIT is robust enough not to care.
|
|
If camera was barely created and there's no info, grab the first frame
to get resolution data.
|
|
|
