ID студента
float64
1k
3.01k
ID треда
int64
1.67k
3.1k
ID реплики в треде
int64
2
16
Имя автора
stringclasses
13 values
Статус автора
stringclasses
2 values
ID предложения
int64
0
6
Текст
stringlengths
3
352
категории эмпатии/антипатии
stringlengths
1
24
виды речевых актов
stringlengths
1
36
модус/диктум
stringclasses
9 values
обоснование (слово, фраза, предложение)
stringlengths
1
47
эмпатия
float64
0.04
1
лингвистическая ошибка (тип)
stringclasses
5 values
текст с примером ошибки
stringclasses
24 values
исправление
stringclasses
23 values
neg
float64
-0.08
0.59
neu
float64
0
1
pos
float64
0
1
compound
float64
-0.75
0.85
3,001
2,501
7
Stepan.Reuk
leader
0
Hello @Sergiy No, at the moment.
ПРИ,ДЕЙ
квалификация
М.Д
hello, at the moment, no
0.5
0.0
1.0
0.0
0
null
null
null
3,001
2,501
7
Stepan.Reuk
leader
0
But this is in our plans
СТОП, ЛЭДМ
прогнозирование
Д
plans, but
0.211364
0.0
1.0
0.0
0
null
null
null
3,001
2,501
8
TAG
leader
0
Hi @Sergiy, thank you for your question.
ПРИ,ВЕЖЛ
благодарность
М
hi, thank you
1
0.0
0.706
0.294
0.3612
null
null
null
3,001
2,501
8
TAG
leader
0
Thinking on possible extensions of Nuitrack object detection module: are you interested in ball detection with a specific/particular appearance such as basketball ball or kind of generic detection of any ball / spherical objects ?
ДВ, НЕУВ
прогнозирование, уточняющий
М
possible
1
0.0
0.926
0.074
0.4019
null
null
null
3,002
2,726
10
Mariya
leader
0
Hi @cinetron!
ПРИ
null
М
hi
1
0.0
1.0
0.0
0
null
null
null
3,002
2,726
10
Mariya
leader
0
At the moment we can’t tell if this is the issue with QT.
ДЕЙ,СТОП
констатация
Д
at the moment, can't
0.5
0.0
1.0
0.0
0
null
null
null
3,002
2,726
10
Mariya
leader
0
We are planning to add support for Ubuntu 20.04 by the end of November.
СТОП
информативное сообщение
Д
null
0.5
0.0
0.828
0.172
0.4019
null
null
null
3,003
1,667
2
olga.kuzminykh
leader
0
Hi Jon, Sorry for a late reply.
ПРИ,ИП
извинение
М
hi, sorry
1
0.178
0.822
0.0
-0.0772
null
null
null
3,003
1,667
2
olga.kuzminykh
leader
0
We’ve checked these tutorial projects and they should work correctly.
-
упрек? констатация? прогнозирование
М
it should work correctly
0.0375
0.0
1.0
0.0
0
null
null
null
3,003
1,667
2
olga.kuzminykh
leader
0
Probably, this issue is connected to the quality of depth map.
НЕУВ,ПРЕДП
прогнозирование
Д
probably
0.5
0.0
1.0
0.0
0
null
null
null
3,003
1,667
2
olga.kuzminykh
leader
0
Please check out our recommendations on sensor placement and environment conditions in our docs 3.
ВЕЖЛ
реквестив
М
please
1
0.0
0.859
0.141
0.3182
null
null
null
3,003
1,667
2
olga.kuzminykh
leader
0
The recommended distance between a user and RealSense D435 is about 1.5 m. Also please check out standard samples, such as nuitrack_sample.
ВЕЖЛ,РЕК
суггестив
М
recommended
1
0.0
0.83
0.17
0.4767
null
null
null
3,003
1,667
2
olga.kuzminykh
leader
0
Is a skeleton tracked correctly?
ДВ
верификативный
Д
null
0.0375
0.0
1.0
0.0
0
null
null
null
3,004
2,735
2
Stepan.Reuk
leader
0
Hello @L0ren4 You mean this app?
ПРИ,ИМП,НБЛАГ
верификативный
М
hello
0.247727
0.0
1.0
0.0
0
null
null
null
3,004
2,735
2
Stepan.Reuk
leader
0
https://apps.apple.com/us/appuitrack/id1476531924 3 This app need for developing apps with Nuitrack functionality https://github.com/3DiViuitrack-sdkree/master/iOS-beta 1 If you run test in this app, you can see rgb stream, depth stream, user tracking (silhouette), skeletal tracking and hand tracking What sensor do you use?
НБЛАГ,НОБР,МВ
объяснение, уточняющий
М.Д
what
0.5
0.0
0.924
0.076
0.4939
null
null
null
3,005
2,767
2
EugeneVasiliev
leader
0
If you need to change the color exactly, then change this lineManual conversion of frames to Unity Texture is outdated and not advisable.
НОБР
объяснение
М
if, then
0.5
0.0
1.0
0.0
0
null
null
null
3,005
2,767
2
EugeneVasiliev
leader
0
Use the extension methods ToTexture(), ToTexture2D(), or ToRenderTexture() for ColorFrame, DepthFrame, or UserFrame.
НБЛАГ,ИМП
прескриптив
Д
use
0.5
0.0
1.0
0.0
0
null
null
null
3,005
2,767
2
EugeneVasiliev
leader
0
Example:..
null
объяснение
Д
null
0.5
0.0
1.0
0.0
0
null
null
null
3,005
2,767
4
EugeneVasiliev
leader
0
You need to change the texture format in the line: For example, on ARGB32 or RGB24 (check the correctness of converting the depth value to the selected format)
НОБР
объяснение
М
need to
1
0.0
0.918
0.082
0.34
null
null
null
3,005
2,767
6
EugeneVasiliev
leader
0
The depthFrame structure is just a matrix, each value of which is a depth value in millimeters.
null
объяснение
Д
null
0.5
0.0
0.758
0.242
0.5859
null
null
null
3,005
2,767
6
EugeneVasiliev
leader
0
You can use depthFrame[i,j] this will be the depth [i, j] of the pixel in millimeters.
РЕК
объяснение
М
null
1
0.0
1.0
0.0
0
null
null
null
3,005
2,767
6
EugeneVasiliev
leader
0
In the example we used: depthFrame[i, j] / 16384f; This means that we divided the depth by 16384 millimeters (approximately ~16 meters) to get an approximation to the normalized values (many sensors work up to 16 meters and we thought that this would be enough).
РЕК
объяснение, прогнозирование
Д
would
1
0.0
0.942
0.058
0.4019
null
null
null
3,005
2,767
6
EugeneVasiliev
leader
0
In fact, the use of this constant is outdated and not correct.
ЛЭКМ
констатация
Д
in fact
0.108333
0.0
1.0
0.0
0
null
null
null
3,005
2,767
6
EugeneVasiliev
leader
0
In the DepthToTexture module (in the NuitrackScripts nested prefab), a variable denoting the maximum depth of the sensor is defined.
ТЕХ
информативное сообщение
Д
null
0.5
0.0
1.0
0.0
0
null
null
null
3,005
2,767
6
EugeneVasiliev
leader
0
It is recommended to use FrameUtils.DepthToTexture.MaxSensorDepth (name space NuitrackSDK.Frame)
РЕК
суггестив
null
recommended
0.5
0.0
0.816
0.184
0.2023
null
null
null
3,005
2,767
8
EugeneVasiliev
leader
0
We recommend using the latest version of NuitrackSDK (v0.35.11).
РЕК,НОБР
суггестив
Д
recommend
1
0.0
0.762
0.238
0.3612
null
null
null
3,005
2,767
8
EugeneVasiliev
leader
0
You need to convert the type correctly.
ТЕХ
суггестив
М
need to
0.5
0.0
1.0
0.0
0
null
null
null
3,005
2,767
8
EugeneVasiliev
leader
0
float depthInMeters = (float)depthFrame[i, j] * 0.001f; Your first example is incorrect, since it will print the last pixel of the depth map.
ЛЭКМ
объявление
Д.М
incorrect
0.04375
0.0
1.0
0.0
0
null
null
null
3,006
2,850
2
irakli
leader
0
Hi @umberto, Which version of Nuitrack are you using?
ОБР,ПРИ,НЕБЛАГ,ДВ
уточняющий
М
hi, which
1
0.0
1.0
0.0
0
null
null
null
3,006
2,850
4
irakli
leader
0
Hi @umberto, I tested Nuitrack on Raspberry Pi 4(Android 10), and the Nuitrack (arm64) test sample loads in 5-6 seconds.
ПРИ,ОБР
объявление
М
hi
0.5
0.0
1.0
0.0
0
null
null
null