-
Notifications
You must be signed in to change notification settings - Fork 5
/
settings.json
85 lines (72 loc) · 4.15 KB
/
settings.json
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
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
{
/* MAIN SETTINGS */
"ip": "192.168.1.110", // Local IP address of the headset to send OSC positions to.
"port": 9000, // OSC port
"debug": true, // Shows debug gui. (May bottleneck FPS, use when having issues)
"fps": 50, // Sets the framerate of the camera.
"model": 1, // Sets the landmark model. 0 = lite, 1 = full, 2 = heavy
/* FILTERING OPTIONS
Each filter value must contain a be null or an object
If it's an object, it must contain a "type" property.
The "type" property must be one of: "Raw", "MovingAverage", "OneEuro".
If set to null, or type is set to raw, no filter will be applied.
The raw data will be used. This may cause jitter.
if type is set to "MovingAverage", the filter will use the
"window_size" property to determine the size of the window.
A greater window_size means less jitter, but more latency.
- "window_size" must be an integer greater than 0. (default: 5)
if type is set to "OneEuro", the filter will use the
"mincutoff", "beta", and "dcutoff" properties to determine
the filter parameters.
- "mincutoff" must be a float greater than 0. (default: 0.05)
- "beta" must be a float greater than 0. (default: 80.0)
- "dcutoff" must be a float greater than 0. (default: 1.0)
*/
// The filter applied to the 2d keypoints detected by the model.
"2d_filter": {
"type": "OneEuro",
"mincutoff": 1,
"beta": 0.00700,
"dcutoff": 1.0
},
// The filter applied to the triangulated 3d keypoints.
"3d_filter": {
"type": "OneEuro",
"mincutoff": 0.05,
"beta": 80.0,
"dcutoff": 1.0
},
/* MULTICAM SETTINGS */
// Settings to use when using more than 2 cameras.
// When multicam_val is smaller than 0, all points with a confidence score
// higher than multicam_val will be used for triangulation.
// If multicam_val is greater than 0, N points with the highest confidence
// score will be used for triangulation.
"multicam_val": 0.625,
/* ADVANCED SETTINGS */
"undistort": true, // Wether to undistort the camera images to accomodate for lens distortion.
"pose_det_min_score": 0.75, // The minimum confidence score for the pose detection model to detect a person.
"pose_lm_min_score": 0.35, // The mininum confidence score for the pose landmark model for a person being in the image.
"refine_landmarks": false, // Wether to refine the landmarks using the heatmap. (doesn't work well currently)
"refine_kernel_size": 7, // The size of the kernel used to refine the keypoints (from the heatmap).
"refine_min_score": 0.5, // The minimum confidence for the heatmap to be used for refining the keypoints.
// Flip detection is an experimental setting which attempts to detect whenever the AI mixes up the left and right side.
// This happens because the face is covered by the VR headset, which causes instability in the pose landmark model.
// The left point of the previous frame gets compared to the right point of the current frame, and vice versa.
// If the distance is smaller than the "flip_detection_max" setting, the points are flipped.
"flip_detection": false,
"flip_detection_max": 10,
"draw_pose": false, // Wether to draw a 3d skeleton visualization.
// Note: The scale multiplier can be found by modifying using the "Real User Height" setting.
// First, adjust the height for your feet to touch the ground (while standing upright).
// Next, start the FBT calibration mode, and adjust the "Real User Height" setting until
// the 3d points align with your in game avatar.
// Finally, divide the first height by the second height, and set the result as the scale multiplier.
"scale_multiplier": 1, // Modify if the 3d points don't align properly with your in game avatar.
"flip_x": true,
"flip_y": false,
"flip_z": false,
"swap_xz": false,
"send_rot": false, // Uses the calculated 3d points for calculation hip rotation. Does not work well due to the AI model!
"owotrack": false // Use owotrack for hip rotation
}