You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: doc/data_processing_single_scene.md
+12-6Lines changed: 12 additions & 6 deletions
Original file line number
Diff line number
Diff line change
@@ -1,12 +1,16 @@
1
1
# Data Pipeline for a Single Scene
2
2
3
-
This outlines how to collect and process data for a single scene. See [here](dataset_organization.md) for how the dataset is organized.
3
+
This outlines how to collect and process data for a single scene. See [here](dataset_organization.md) for how the dataset is organized. The steps here are split across code in two repos.
4
+
-[spartan](https://github.com/RobotLocomotion/spartan) handles the raw data collection and tsdf fusion.
5
+
- pdc handles change detection and rendering.
4
6
5
-
## Capture Raw data with Kuka
7
+
## Spartan
8
+
9
+
### Capture Raw data with Kuka
6
10
7
11
The quick version of raw data collection currently is:
8
12
9
-
#### Human-moved objects
13
+
#####Human-moved objects
10
14
11
15
1. Start Kuka, run Position control
12
16
2.`kip` (shortcut for Kuka Iiwa Procman) then in procman:
@@ -23,7 +27,7 @@ cd ~/spartan/src/catkin_projects/fusion_server/scripts
23
27
2.`./capture_scene_client.py`
24
28
3. This will create a new folder with the current date (e.g. `2018-04-07-20-23-56`) and the `raw/fusion.bag` file as in the folder structure above.
25
29
26
-
#### Autonomous robot-moved objects
30
+
#####Autonomous robot-moved objects
27
31
28
32
1. Start Kuka, run Position control
29
33
2.`kip` (shortcut for Kuka Iiwa Procman) then in procman:
@@ -33,7 +37,7 @@ cd ~/spartan/src/catkin_projects/fusion_server/scripts
33
37
4. Run Director
34
38
3. In Director terminal (f8), enter: `graspSupervisor.testInteractionLoop()`
35
39
36
-
## TSDF Fusion
40
+
###TSDF Fusion
37
41
This is done in `spartan`. Navigate to `spartan/src/catkin_projects/fusion_server/scripts`. With `log_dir` set to the directory of your log, i.e. the full path to `2018-04-07-20-23-56` run
38
42
39
43
```
@@ -50,7 +54,9 @@ This will
50
54
3. Convert the tsdf fusion to a mesh and save it as `processed/fusion_mesh.ply`
51
55
4. Downsample the images in `processed/images` and only keep those with poses that are sufficiently different.
52
56
53
-
## Change Detection and Depth Image Rendering
57
+
## PDC
58
+
59
+
### Change Detection and Depth Image Rendering
54
60
This is done in `pytorch-dense-correspondence`. In `pdc`
0 commit comments