Brandon Thomas
9fedbcd967
refactor frontend into directories by mode
2020-05-27 00:23:13 -04:00
Brandon Thomas
3ce772cd80
Modal test frontend can hit /speak or /advanced_tts
2020-05-27 00:17:10 -04:00
Brandon Thomas
ac1a9281d4
CORS, logger
2020-05-26 22:46:22 -04:00
Brandon Thomas
d4b1fa93b1
Add CORS allow list
2020-05-26 22:21:09 -04:00
Brandon Thomas
450c4ebe0a
Add /speak and /speakers endpoints
2020-05-26 20:57:40 -04:00
Brandon Thomas
0652199e8a
ignore models and wav files
2020-05-26 20:37:55 -04:00
Brandon Thomas
a20a7259b5
beef up module config: shared base dir, speakers
2020-05-26 20:36:08 -04:00
Brandon Thomas
805bad97c3
server code cleanup, cache model configs
2020-05-26 06:15:59 -04:00
Brandon Thomas
729bdefdbc
Updated deployment
2020-05-26 05:37:25 -04:00
Brandon Thomas
dafe174dc3
Repair dockerfile: needs libtorch 1.5; update symlinks
2020-05-26 02:52:05 -04:00
Brandon Thomas
d3a0a30884
tts_service: specify model config file by env var
2020-05-26 01:15:58 -04:00
Brandon Thomas
0a70b326e7
Kind of stuck, now. Rendering red and blue point clouds
...
- Color information from second camera only gets through if first camera
is disabled.
- How do I print BOTH point clouds from uniform data?
2020-05-18 04:47:45 -04:00
Brandon Thomas
6141c453fd
two point clouds translated on the x-axis
2020-05-18 01:07:33 -04:00
Brandon Thomas
d7d52485d7
two point clouds! (gl_VertexID-interleaved rendering)
2020-05-18 01:01:15 -04:00
Brandon Thomas
983d301554
getting somewhere, but this is confusing.
...
Changing textures in the point_cloud_renderer GLSL to be either index 0
or 1 (pointCloudTexture0 or pointCloudTexture1, each corresponding to a
camera) to the opposite of the skip block in the loop on line 611 will
show a perfect shaded image from the non-skipped camera.
This suggests that both cameras are sending texture and color data.
However...
Trying to do a linear combination from non-skipped color data is WEIRD
though, and not as I'd expect.
2020-05-18 00:47:55 -04:00
Brandon Thomas
56b4271295
Something interesting going on.
...
- Both inputs are some combination of colors
- Covering one camera leads to a black image
- Covering another camera leads to a red-shaded image
One of the cameras must not be wired up correctly.
2020-05-18 00:11:09 -04:00
Brandon Thomas
cfb4f8d6dd
Handle multi-cam vertex array object buffer for color (doesn't work yet)
...
The code works to load multiple VAOs and buffers, but the camera data
isn't properly segregated yet. Need to figure out what to bind
2020-05-17 23:10:50 -04:00
Brandon Thomas
ce64b80fed
upgrade tch.rs 0.1.3 -> 0.1.7 (still segfaults)
2020-05-17 19:47:54 -04:00
Brandon Thomas
ad0ca149f8
I *think* this works now. See binding=0,1 in GLSL
...
Note: we're currently looking at pointCloudTexture1, which is a
different depth camera from the color data
2020-05-17 05:55:58 -04:00
Brandon Thomas
ed74711f80
switch between camera for point cloud renders
2020-05-17 04:10:29 -04:00
Brandon Thomas
a882e89b1c
show both converted depth xyz textures!
2020-05-17 03:59:25 -04:00
Brandon Thomas
83d7534b26
Visualizer now switches between multiple cameras (one at a time)
2020-05-17 03:37:34 -04:00
Brandon Thomas
99a9a2f966
Incremental support for multiple cameras: PointCloudVisualizer
...
Lots of internal use of Vec<T>, but still tied to 0-indexed items
Code quality --
Still compiles and runs
2020-05-17 03:06:42 -04:00
Brandon Thomas
d09aab5743
Incremental support for multiple cameras: Multiple PointCloudConverter
...
Code quality --
Suport ++
2020-05-17 02:40:57 -04:00
Brandon Thomas
83cee2c4f2
make more readable by sending basenames to frontend
2020-05-15 03:00:49 -04:00
Brandon Thomas
9772cabd99
Model picker! Updated frontend and backend
...
This allows the frontend user to choose which melgan and tacotron are
used at runtime without restarting the server. It could use UI
improvement, but it works.
The available models are configured with a toml config file loaded by
the backend.
2020-05-15 01:20:52 -04:00
Brandon Thomas
f5cf28a589
upgrade sentence.rs
2020-05-11 02:21:28 -04:00
Brandon Thomas
b3432795e5
Better space insertion, handling of punctuation
2020-05-11 00:30:10 -04:00
Brandon Thomas
6139bc6b3d
Add sentence/numerics.rs for tokenization, numbers
2020-05-10 22:55:15 -04:00
Brandon Thomas
7b2eff1387
uncommitted changes to frontend
2020-05-10 22:42:07 -04:00
Brandon Thomas
8625daa59d
Add frontend index.html
2020-05-10 22:39:00 -04:00
Brandon Thomas
d2f517b092
doc commit
2020-05-06 20:58:38 -04:00
Brandon Thomas
33bf87a1e4
uncommitted changes
2020-05-06 20:58:05 -04:00
Brandon Thomas
d18efaaefc
make enabling webcam easier
2020-05-05 05:15:12 -04:00
Brandon Thomas
83954a0742
update so we don't have to specify secondary devices
2020-05-05 05:12:31 -04:00
Brandon Thomas
c83bf467c0
synchronized capture! (still only reading one image)
2020-05-04 05:38:41 -04:00
Brandon Thomas
5bb60032c3
multi-device capturer (starts multiple cameras in order)
2020-05-04 04:44:38 -04:00
Brandon Thomas
9081a3da26
refactor: organize kinect and opengl modules
2020-05-04 03:28:41 -04:00
Brandon Thomas
a4d16224c2
refactor (move arcball to gui code)
2020-05-04 02:10:56 -04:00
Brandon Thomas
2a0cb25a8f
no more SDL arcball for now
2020-05-04 02:07:02 -04:00
Brandon Thomas
a06afdde38
handle zoom only within imgui window
2020-05-04 02:05:11 -04:00
Brandon Thomas
c2c9dbceed
animation hacks
2020-05-03 22:22:33 -04:00
Brandon Thomas
52338b9bcb
animation
2020-05-01 07:25:05 -04:00
Brandon Thomas
95fec69018
stupid experimentally determined camera zero :(
2020-05-01 07:13:37 -04:00
Brandon Thomas
67a10b0892
custom perspective matrix based on k4a
2020-05-01 06:07:32 -04:00
Brandon Thomas
991f1b8aaa
clear color, rgbaf32 type
2020-05-01 05:16:33 -04:00
Brandon Thomas
ba5366d65e
clean up some unused imports
2020-05-01 04:53:29 -04:00
Brandon Thomas
65f1d5b980
ran cargo fix and then CLion import optimizer
2020-05-01 04:40:21 -04:00
Brandon Thomas
21d40a9fe9
about to run cargo fix
2020-05-01 04:28:54 -04:00
Brandon Thomas
790e0a9748
move old code (preserved since I'm still new to this)
2020-05-01 04:25:40 -04:00
Brandon Thomas