Clean up data

If your visual assessment and Vicon Shogun Post's diagnostic tools have revealed issues with your recorded data (see Check data quality), you can use its cleanup tools to correct mislabels and fill any gaps.

See also the related Vicon videos: 5 - Shogun Post – Labeling Data and 6 - Shogun Post - Marker Editing.

Note that each time you finish cleaning up a range, you must check its solve (see Solve during cleanup).

Find and fix bad data

To help you identify and fix bad data, you can use the Marker Editing panel.

To find and remove bad data:

1. To open theMarker Editing panel, in the Processing menu, click Marker Editing (or click Marker Editing on the Processing tab of the ribbon).
2. In the Remove Bad Data section, go to the Find Bad Data button and notice the options:
   
   • Threshold Allowable deviation
   • Cut-Off Data filter. Decreasing this value filters the data more heavily.
   • Sensitivity Amplifies the effect of the Cut-Off filter.

     

3. Click Find Bad Data.
4. Experiment with these values to find out what works best for your data.
5. When you have identified the bad data, remove it by clicking the appropriate Cut button, and apply a fill (see Gap-filling options).

Find and fix noise

Noise can be easily identified in a Graph view:

It also produces flickering as you play through a take in the 3D Scene view. You can fix it with the filtering options in the Marker Editing panel.

To fix noisy data:

1. In a Graph view, ALT+drag to select it.

   

2. At the bottom of the Marker Editing panel, expand the Filter section, and select whether filtering applies to Ranges, as above, or to Selected Keys.
3. If you're not sure how much filtering to apply to your data, keep the default settings (Cut Off: 0.3 and Threshold: 15) and click Apply. You can reapply this as many times as required.

   

4. In the Graph view and 3D Scene view, check that the trajectory is now smooth over the selected range.

   Tip: To display a tooltip for any of the controls, hover the mouse pointer over the relevant control.

Ensure smooth start and end to filtered trajectories

When you apply filtering, the Smooth In/Out option in the Marker Editing panel provides smooth blending at the start and end of the filtered section of a trajectory curve.

The following example shows a Graph view of a trajectory curve that requires smoothing.

With Smooth In/Out cleared, when filtering is applied, the curve is made smoother, but the start and end do not take into account what is on either side of the smoothed section, resulting in a straight line, with an unwanted angle at the end.

For details, see the following steps or watch Vicon Shogun 1.3 Post Tutorial - Filter in/out Smoothing on YouTube.

To give a smoother start and end to the filtered section of a trajectory:

1. In the Marker Editing panel, expand the Filter section and select Smooth In/Out.

2. If you want to change how much of the trajectory is affected on either side of the gap, edit the value of the adjacent field. The default is 0.2 (20%). Smoothing works by blending between the original and filtered result by progressively more or less, across the smoothing range.

   

   The filtered section is smoothed at the start and end.

Use the Labeling panel

You can fix labeling issues using the tools in the Labeling panel. The following steps introduce the main components of the labeling panel. To go straight to examples of how to use it to fix labeling issues, see Fix common labeling issues.

To use the Labeling panel:

1. On the Processing tab on ribbon, click Labeling.
2. In the toolbar at the top of the Shogun Post window, ensure that the required subject is selected in the Current Subject list.

   Tip: By default, the selection in the Current Subject list at the top (middle) of the Shogun Post window determines which subject to label. If All is selected in the Current Subject list, the labeler uses the last subject it was set to. If you want to select a subject different from that specified in the Current Subject list, in the Labeling panel, clear the Use Current Subject box and select the required subject from the Subject list at the left of the check box.

   On the left of the Labeling panel, a list of labels for the selected subject is displayed. The color variations indicate marker issues:
   • Yellow: Mislabels or missing labels (the depth of the color indicates the severity of the issue, eg, more or fewer gaps)
   • Red: Labels for this marker are missing from the current frame

   

3. To display a 3D representation of your labels for the current subject, click the 3D button at the top of the Labeling panel. The 3D view helps you to quickly identify where the markers should be. You can drag and drop labels from the 3D Labeling view to the 3D Scene view pane (and vice versa).

   

   You can use the usual mouse actions (click and drag, right-click and drag, left- and right-click and drag) in the 3D view, in the same way as in a 3D Scene view.
4. At the top of the Labeling panel, in the Manual Labeling Options section, select the Mode option (Select or Label), which affects the way in which you select and label markers.
   
   • Select Click a marker on your subject in a 3D Scene view and then click a label name in the list in the Labeling panel.
   • Label Click name of the label in the marker list and then click the required marker on the subject in the 3D Scene view.

   Tip: To quickly switch between labeling and select modes when labeling a subject, you can use the default hot key (L) that duplicates the functionality of the Label and Select buttons at the top of the Labeling panel.

5. In the Direction line, choose to label forward (through time) or backward .
   You can also select both options (ie, label both forward and backward), but to help you avoid confusion, at least initially, choose either backward or forward and use only that option.

6. From the Type options, select the way in which labeling will be applied:

   

   • Whole Labels entire trajectory.
   • Fragment Labels the trajectory that intercepts the current frame.
   • Cliff Labels the current frame and continues until a specified value is encountered, which stops the labeling (see the text below in theLabeling panel, for example, the default is to skip gaps smaller than 5 frames and stop labeling at cliffs that are larger than 50 mm).
   • Ranges Lets you select an area on your graph or timeline and label only the selected time range.

In the Manual Labeling Tools section in the middle of the Labeling panel, you can correct swaps, unlabel information, and unlabel markers.

In the Semi-Automated Labelers section, you can access the Velocity Labeler, which is normally used after automated labeling, on a partially labeled take. You can use it to correct labeling where the path of a single marker consists of multiple trajectories that are consecutive in time with a small gap in between where the marker is unlabeled over part of its trajectory. The Velocity Label option is useful when a marker has been labeled for a range of time, and then becomes unlabeled, yet going forward or backward in time there are multiple trajectories that do not have a large gap between them and are all the same marker. It is best suited for cases when unlabeled trajectories are not many frames away from the labeled marker and the velocity of the marker around the end of the labeled marker and the start of the unlabeled trajectory is fairly constant.

Fix common labeling issues

The following basic procedures are just a few of the ways in which you can use Shogun Post to clean up data.

• Identify issues
• Correct a swap
• Label an unlabeled marker
• View and select gaps
• Auto-fill with intelligent rigid fill
• Manually fill gaps

See also:

Vicon Shogun 1.3 Post Tutorial - Gap List & Auto Rigid Fill on YouTube.

Identify issues

1. To help you identify issues, in the 3D Scene view, click View Filters, and ensure that in the Scene section, Missing Markers is selected. Also ensure that in the Subjects section, Label is selected and that in the L(abeling) column of the views matrix, Constraints is selected.

   

2. In 3D Scene view, scrub through the whole take (drag the current time indicator along the timeline, or for finer control, press the A or S key), noting times when markers are likely to be occluded (sitting, crouching, subjects interacting, etc).
3. In the marker list in the Labeling panel, note any missing (red) markers and watch for ‘popping’ or moving in the 3D Scene view. Look for any swapped markers (for an example, see the Vicon video 5 - Shogun Post – Labeling Data, which shows an example of a swap of knee and heel markers). Missing markers are clearly visible, highlighted in red. However, after you’ve identified the erroneous markers, it may help with labeling to clear Missing Markers and Labeling Constraints.

Correct a swap

1. Ensure the Labeling panel is displayed.

2. Split the 3D Scene view (at the top right of the Shogun Post window, click the vertical split button ) and in one of the panes, change to a Graph view, so that you can now see a 3D Scene view, a Graph view and the Labeling panel.
   
   
   
   

3. In the Labeling panel, ensure Select mode, Forward direction (>), and Cliff are selected.
4. In the 3D Scene view, select a marker that is incorrectly labeled and in the Graph view, zoom in (right-click and drag) and go to the start of the where the swap happens (this should be identifiable on the graph by a sharp change).

   

5. From the position of the marker in the 3D Scene view, decide which is the correct marker and click it in the 3D view.
   In the marker list, its name is displayed in heavier text, and at the top left of the 3D Scene view, the Selection text displays the name of the selected marker.

   

6. Click the wrongly labeled marker in the 3D Scene view and then click the correct marker name in the Labeling panel list.
   The marker is now correctly labeled. To check, scrub back and forward in the 3D Scene view and check that the marker is now behaving correctly.

However, you now have a previously incorrectly labeled marker that is unlabeled from this point forward.

Label an unlabeled marker

1. In the 3D Scene view, click on the unlabeled marker to select it. Scrub back and forward to identify where it is correctly labeled. This is the point at which it is correctly colored in the 3D Scene view and the Labeling panel marker list. You can also observe this on the Graph.

   

2. To label a marker using the tools in the Labeling panel, do one of the following, depending on your chosen Mode:
   If you prefer to label using Label mode:
   
   1. In the Manual Labeling Options section, in the Mode line, click Label.
   2. In the marker list in the Labeling panel, click the name of the marker and then in the 3D Scene view, click the marker that is currently unlabeled.
   Or, if you prefer to label by dragging, using Select mode:
   
   1. In the Manual Labeling Options section, in the Mode line, ensure Select mode is selected.
   2. In the Labeling panel, click to select the required marker in the 3D view.
   3. SHIFT+CTRL then click+drag from the selected marker in the Labeling panel to the unlabeled marker in the 3D Scene view.
3. As before, to check the labeling, scrub back and forward in the 3D Scene view and check that the marker is now behaving correctly.

After you have corrected any swaps, you may want to use Shogun Post's diagnostic tools to help to identify gaps (see Check data quality).

View and select gaps

You can view all the gaps in your scene in the Marker Editing panel, in the Fill Gaps section, by looking at the Gap List. (If you can't see the list, at the top of the Marker Editing panel, click the Gap List button.)

To make it easy to find the longest gaps, click the Length column heading to rearrange the list, with the longest gap at the top.

When you select a gap in the list, it is automatically displayed in the Graph view, enabling you to edit it.

You can then decide the best approach for filling the gaps:

• As a first step, try automatically filling as many gaps as possible (see Auto-fill with intelligent rigid fill).
• If your scene still contains gaps after using the automatic fill, or if auto-fill is unsuitable, try manually filling the gaps (see Manually fill gaps).

  Tip: To undo unwanted gap-filling, click Undo on the quick access toolbar .

Auto-fill with intelligent rigid fill

Shogun Post enables you to automatically fill gaps using a rigid fill operation. This looks at all the markers in your scene and then compares them against the marker you are trying to fill. It then uses a combination of similarly moving markers to fill the gap. Finally, it checks the fill to make sure it looks correct and if not, it chooses another set of markers.

This process is also available via scripting, using the selectMarkersForRigidFill command to select the markers and the autoFillGaps command to fill them. For information on these commands, see HSL scripting with Vicon Shogun.

To auto-fill a selected marker:

The following procedure for rigid gap-filling is semi-automated, in that you choose a marker for the fill:

1. In the Marker Editing panel, expand the Fill Gaps section and in the Fill Rigid area, select the Auto-select markers for fill check box.
   To change the default options for auto-selection, click the Show Options button next to the check box. The options are:
   
   • Max Deviation: Specifies in mm how rigid the set of markers used to fill must be, as the maximum deviation in distance between all pairs in the set. For example, a value of 60 enables the markers to flex in rigidity by 6 cm.
   • Max Distance: Specifies in mm how far away the candidate can be from the selected marker. The default of900 is just under 3 feet (1 m), because a greater distance is likely to be another subject or body part, which is unlikely to move rigidly during the gap.
   • Percent Time Rigid: Specifies the percentage of time over the adjacent fragment ranges that the selected marker must be rigid with the chosen set. The default of 0.75 means 75% of the time.
2. Select a marker with a gap.
3. Ensure the current time is in the gap.
4. Click Fill Rigid.
   Gaps are filled using the intelligent rigid fill.

To auto-fill selected markers or all markers:

In addition to the semi-automated procedure described above, you can also automatically fill all gaps on either all markers, or only those currently selected. To do this:

1. Depending on which markers you want to fill:
   • Only selected markers. Ensure you have selected the required markers.
   • All markers. Ensure no markers are selected.
2. In the Marker Editing panel, expand the Fill Gaps section and in the Fill Rigid area, select the Current Frame, Selected Ranges or All Time option.
3. Select Auto-select markers for fill and ensure the options are as required (see above).
4. Click Fill Rigid.
   All gaps on the selected markers, or on all markers if no markers are selected, are filled over the play range, or selected ranges.
   Note that in the case of all markers, only labeled markers are filled, though unlabeled markers can be auto-selected for filling the labeled marker.
   
   

You can run automated gap-filling multiple times to achieve the required results, in particular when:

• After first run, it either made an unwanted fill or wasn't able to fill a gap due to a mislabel. Fix the mislabel and then run it again to fill the remaining gaps.
• After first run, it either made an unwanted fill or wasn't able to fill a gap due to the gap being too complex. Fill the gap manually, then run automated gap-filling again to fill any remaining gaps. These can now can be filled due to more data being available.
• After first run, some gaps remain. Loosen the rigidity settings (see Note above) and run it again.

Manually fill gaps

You fill gaps using the Marker Editing panel, normally in conjunction with the Data Health view and/or a Graph view. The Marker Editing panel contains all the tools necessary to fill gaps, alter trajectory keys and filter your data.

For example, you might first notice a gap from the display on the time bar Issues map, then note which marker is affected in the 3D Scene view:

1. To quickly gap-fill the selected marker, on the time bar double-click to move the current time indicator to the relevant frame.
   The affected marker is automatically highlighted in the 3D Scene view (you can right-click and drag to zoom in further to check the marker).

2. To open a Data Health view and a Graph view, as well as a 3D Scene view, split the screen by clicking the Three Views Split Left button at the top of the Shogun Post window and change the views in the new panes.

3. If the Marker Editing panel is not already open, on the Processing tab on the ribbon, click Marker Editing.
4. In theData Health view, find the relevant marker's line, which shows a gray rectangle that represents the gap.

   

5. In the Data Health view, double-click to select the range of the gap.

   

6. In the Marker Editing panel, expand the Fill Gaps section. In the Fill Using Interpolation area, select Selected Ranges (to fill only the current selection), and then click Fill Using Interpolation.

   

   This looks at the keys on either side of the gap and fills it by interpolating between them. It works well for small gaps.
7. In the 3D Scene view and Graph view, check that your fill has given the required result.

   

8. If the fill does not give an appropriate result, you can click Undo on the quick access toolbar , and try other gap-filling options in the Marker Editing panel (see Options for gap filling).

Options for manual gap-filling

The appropriate option to use depends on the type of gap that you want to fill.

Short gaps

1. Manually select a short gap by double-clicking in a Graph view orData Health view, and in the Marker Editing panel, expand the Fill Gaps section and click Fill Using Interpolation.
   This looks at data before and after the gap and uses a spline fill to interpolate between them.
2. Check that the result looks reasonable in a 3D Scene view and Graph view.

Sequence of short gaps

1. For a sequence of small gaps with a smooth line running through the gap in the Graph view, ALT+drag to select the affected data.

   

2. In the Fill Using Interpolation area, select Selected Ranges, and then click Fill Using Interpolation.

   

Non-linear motion

For non-linear motion, for example, where rotational movement is involved, as interpolation does not account for this kind of motion, a different approach is needed. If a missing marker is part of set of markers that are rigid (ie, remain in the same relation to each other, for example, a pelvis), you can use three or more markers in the rigid object (three in addition to the marker that has the gap is recommended) to help fill in the data for the missing marker.

1. To select the range of the gap, double-click in a Graph view orData Health view, then select (CTRL+click) the three other markers in the rigid object.

   

2. In the Marker Editing panel, expand the Fill Gaps section.
3. In the Fill Rigid area, select Selected Ranges and click Fill Rigid.
   The pattern of the present markers is copied onto the missing marker.
4. As usual, check in the Graph view and in a 3D Scene view that the fill looks sensible.

Other rigid objects you can use in default marker set include hands, forearm, upper arm, shoulders, thorax, and head.

Non-linear motion where rigid fill impossible

In cases where the motion is non-linear, but a rigid fill is not possible due to all the related markers being absent, you can use the Fill Using Constraints option. This uses the labeling or solving skeleton to produce the fill. It requires a skeleton be present, the skeleton to be constrained by markers, and enough markers to have data that the solve is fairly good. The marker being filled must be constrained to the skeleton because it is the constraint offset that is used to determine where the marker should be during the fill.

Before using this type of fill, check that the bones and missing markers (by default displayed in red in the 3D Scene view), are in sensible locations.

Processing during cleanup

Note that the previous steps for fixing data issues assume that your existing .mcp data is largely of acceptable quality. However, in some circumstances (for example when you are trying to produce better quality data), rather than persisting in trying to fix a problematic .mcp file, you may need to clear the scene and start from the .x2d file, or unlabel all the data first. In this case, do not just open the Processing panel and click Reconstruct or Label or run other Combined Processing operations as this will not fix problems with the underlying data.

Clear existing data

To remove all unlabeled data and clear existing labels and solves, at the top of the Processing panel, select the Reset Scene check box. You can then run Reconstruct, etc. Note that you can't undo the reset.

To remove existing labels from the current file, on the Labeling tab of the Processing panel, select Clear Existing Labels and then run Labeling as required.

To remove occlusion fixing from your data, in the Marker Editing panel, expand the Restore section and select the required option.

Solve during cleanup

During the cleanup operation, each time you finish cleaning up a range, check its solve. To do this:

1. Ensure the range you cleaned up is selected.
2. On the Processing tab on the ribbon, click the arrow on the Solve Solving button.
3. Click Solve Solving Ranges. The selected range is solved, enabling you to quickly check that your cleanup has been successful, without having to solve the whole take.
4. Proceed to clean up the next range that contains mislabels, gaps or noisy markers, etc. After you have completed all the cleanup required, finish by solving the entire take. For more information see Solve the data.