Okay, I put in 35mm for the “image sensor size” and put in 2.3 mm for “focal length” and set the solver to refine both the “sensor size” and the “focal length”.
I got a solve error of about 82.
Also, there is a blue line on the bottom of the video that tells you which frames have been cached. For me, half of that line is red. There is also an error message in the “info” window that says “some of the data failed to reconstruct”. In the console window, there are multiple messages saying “no camera for frame x” for about half of the frames in my video.
Ok well that is not what I suggested. What I was trying to say was, leave the sensor size at 35mm in Blender but use the focal length for the 35mm equivalent. At the widest that is 36mm.
You were mixing the real sensor lens length with the equivalent sensor lens length.
Also it sounds like you need more tracking markers for the rest of the shot. You must have 8 at any time throughout the shot (IIRC) other wise at least 8 during solve time (the frame range you define for greatest parallax change).
3Point, it sounds like he may not be familiar with 35mm photography, or converting from one format to another.
Anezina, my Panasonic camcorder doesn’t record EXIF data for video, but it does record it for still photos. So what I do is to shoot a still photo at the same focal length that I’m shooting my video at. Then I just check the EXIF data for the still image and use that in the tracker’s camera settings. (highlighted in the pic below. Note that mine gives both actual and 35mm equivalent focal lengths.)
Also, it looks like your camera only shoots video at 640x480. You might want to invest a couple hundred bucks in a camera that shoots HD. The higher resolution means the spots that you’re tracking will be more detailed and won’t be so pixelated. It should result in better tracks.
I put in “35 mm” for the image sensor, and “36 mm” for the focal length.
I also checked to make sure I had at least 8 active markers at any given time.
When I solved, I still got high errors. I then set the solver to refine the focal length and hit the “camera motion” (ie solve) button several times. The errors varied wildly. 58, 26, 42, 97, 8, 17, … I got lucky at one point when the error dipped to below 1.0 I (optimistically) tried to import everything into the 3D , but then the video wasn’t coming up in the background when I switched to the 3D camera view.
Steve S, my camera is only 640 x 480, however I had little problems putting down the markers. There were two or three that I had to help along, but the other markers tracked nicely. I don’t know how to do your EXIF suggestion since my camera doesn’t export images in EXIF format.
Do you have sufficient parallax for a good solution? You want to set the keyframe settings to a part of the video where there’s the most parallax.
Can you post the original video to Youtube so we can review it?
The problem was that my camera motion was too complex. Originally, I had the camera pan the whole length of a porch from the ‘street’. Then the camera went onto the porch and rotated around, filming the whole porch.
I solved the problem by breaking the video into separate parts. The rotating is now a separate video.
Correct me if I am wrong, but for camera tracking to work, the camera motion should be simple and short ?
Thanks for your suggestions. The suggested focal lenghts and image sensor widths steered me in the right direction.
After I finished solving, I clicked “set as background” (in the Movie Clip Editor’s “reconstruction” menu) in order to bring the video clip into the 3D viewport so that I can start making the 3D object that I want to import into the video.
Unfortunately, the video does not appear in the 3D viewport. Even when I am in camera view.
Also, when I render, my 3D object appears black and semi-transparent. I did click “set up tracking scene” as instructed by the tutorials.