First look at PFTrack
Tree view- node based trees with pipeline.
Middle rectangle- canvas giving ability to view what you are looking at.
First things first- create a new project. Do this by selecting create button under the canvas window.
If you click create again- the screen will change to the below image.
On the LHS we still have the tree view, the middle is the media bins and then the navigator. The navigator moves through the pipeline allowing you to chose where you pick the footage from. N.B. PFTrack does not have a save button but does automatically save the file for you (yay).
Accessing PFTrack’s Preferences
Setting up the application to run best with the machine. To access the menu to do this click the spanner/ clog icon at the top RH corner.
In the cache option- once the bar reaches 100%, click the clear cache button. If you have a computer with high specs, you can up the cache to 4GB.
In the General, you can tell the software how many CPU’s to use. If you go to USE- select the drop to chose, you can type in for e.g. 1/8 to allocate how many you want the computer to use. The next is the undo box- setting how many undos you can use- set to 20.
You can also edit the scene units- i.e. in centimetres, feet etc.
In the cinema tog you can change the view of the trackers. You can change the tracker colour and the shape.
In export- there is the having path. Where a project is set there is a sub folder named export. You can also alter the type of file (JPEG, TIF etc).
Using the Media Admin Menu
Click the two page icon on the canvas viewport. Then find the footage- click and drag to the tree port. The settings for the footage are loaded into the window below.
N.B. click the C on the timeline to cache the footage to the timeline. In the settings cache menu- click on the cache menu and see what the percentage is. If it is at 99%- clear the cache.
Next look at the media admin menu. You will see two loaded clips- the bottom being the original and the top being the cached one. If you look to the RHS you will see a green lit ‘OK.’ This ok means that the software can locate the footage and read it.
Going back to the footage- the tutorial addressed two problems that are common in tracks. The first being lens distortion. This causes problems as tracker points are put onto the distortions- giving false tracking information. The second problem is that the picture is a bit blurred. Blurry footage gets rid of trackers and contrast- things that they like to be attached to. This is fixed with corrections inside the tracker.
Click create- showing all the nodes in the editor. Click on the image manipulation node. The tutorial demonstrated that if in the filter option in this node- if you tick it and try to fix the sharpen radius, it removes the cache from the footage. The colour can also be turned on to alter. Which channel is the cleanest to use? The RGB channels are accessed through the three dots under the Tree view.
In the example above the green is the cleanest and the blue is the nosiest. Therefore, the greens and reds need upped more, and the blues slightly. The next step it to also up the saturation and reducing the gamma. To reset any of these, click the R beside the levels.
Next, go back to the create and go to the undistort node. This node will allow you to match the lines up with lines in the footage. The screen will appear as below.
To show the computer the level of distortion- the example used here was the lamppost. Drawing a line at the top and bottom of the post, and then putting more in the middle to allow the bend, will give the amount of alteration needed.
Once this is done, click the solve button and the computer will solve them. Once this is done, go ahead and cache the footage again.
Tracking Footage Manually and automatically
Going back to the create option to show the node menu- we have two options. Using an auto track or a manual track. The different settings are as follows;
Search mode- better accuracy
Deformation- scale and rotation (due to being a hand held camera)
Consistency- free camera (as it is handheld)
Failure Threshold- setting a lower value (i.e. 0.5) for problematic footage is better- any tracker outside 0.5 will be a bad translation and will not be made. Blur image also helps if there is too much noise. Clicking auto track allows these to be tracked. The software tracks forward then backwards to average out the values of the trackers.
Auto tracker is not good in the fact that is a soft track- trackers do not stay true to their point value.
In this case- user track will have to be used.
Solving The 3D Camera
Click on the camera solver node.
We can tell software which is a hard tracker or not (true to the origin). If the camera is shaky you are best to set the translation and rotation to medium smooth. After this, select solve all.
After go to the errors tab. Any tracker over 1 has high error margins. Select trim.
Go to the 2H button to see how the solve will look in 3D space- holding command will allow you to rotate around this.
Orienting our Scene and Test Objects
As scene in last tutorial the camera with the point cloud is not sitting on the ground. The ground (grid) is floating and the horizon line of the grid does not match that of the footage.
Using a orient scene node to adjust the scene to the footage. In the edit mode box on the RHS- you can change the different orientation modes (rotate, resize, translate etc).
Once matched up- a test object can be used to test the legit-ness of the track. The tutorial used a thumb tack as an example. These are added to the trackers by clicking on the tracker, then clicking on the marker. After one, the duplicate button can be used to mark other trackers.
Exporting the Scene to Maya
To export create an export node then select the type of file. Before exporting ensure in the objects mode the thumbtacks are taken off so they don’t go too. Finally select export to export the camera.
When opening the scene in maya it is important t0 ensure the size of the scene matches that of the track (cm, m, ft etc). Once the camera is brought into maya, increase the tracker size to 0.5 to make them more visible. In this example the trackers do not match to the original footage, so the altered footage from PFTrack also has to be uploaded to the camera as a camera plane.