Brox 密集光流跟踪器错误“EFilterIncompleteSize”
我正在使用 Thomas Brox 提供的 Linux 64 位非 GPU 可执行密集光流跟踪器（可在此处获取：<http://lmb.informatik.uni-freiburg.de/people/brox/code.html >)。我已按照此处的说明进行操作，并正确设置了一个 .bmf 文件，其中列出了我自己的用于跟踪的自定义图像。我已经安装了此可执行文件中包含的库文件。一切似乎工作正常，跟踪可执行文件开始处理我的帧，就像处理其中包含的示例帧一样。
然而，在开始处理后不久，我收到此错误：
Exception EFilterIncompatibleSize: Initial container size: 2232 Resulting container size: 2592
terminate called after throwing an instance of 'EFilterIncompatibleSize'
Aborted
我在留言板帖子或在线其他任何地方找不到对此错误的任何引用。有其他人在您自己的图像文件上成功使用 Brox 的跟踪器吗？如果是这样，您能解释一下这个错误吗？
I am using the Linux 64-bit non-GPU executable dense optical flow tracker provided by Thomas Brox (available here: < http://lmb.informatik.uni-freiburg.de/people/brox/code.html >). I've followed the instructions there and correctly set up a .bmf file listing my own custom images for tracking. I have installed the library file included with this executable. Things seem to be working correctly, and the tracking executable starts to process my frames much like it does for the example frames included with it.
However, shortly after starting to process, I get this error:
Exception EFilterIncompatibleSize: Initial container size: 2232 Resulting container size: 2592
terminate called after throwing an instance of 'EFilterIncompatibleSize'
Aborted
I can't find any references to this error on message board posts or anywhere else online. Has anyone else used Brox's tracker successfully on your own image files? If so, can you shed any light on this error?
如果你对这篇内容有疑问，欢迎到本站社区发帖提问 参与讨论，获取更多帮助，或者扫码二维码加入 Web 技术交流群。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱，如果其他用户或者作者回复了您的评论，将不能在第一时间通知您！
发布评论
评论（1）
经过更多挖掘和联系作者后，发现这个错误是由于向跟踪器提供不同尺寸的图像而引起的。容器大小与图像的像素尺寸有关。就我而言，我相信我的所有图像都具有相同的尺寸，因为它们是裁剪程序的输出。然而，一些被自动裁剪的区域非常接近图像边界，因此裁剪会截断它们，导致裁剪区域各处缩小几个像素。如果您遇到此错误，这是我首先要检查的事情。
After some more digging and contacting the author, it turns out that this error is caused when images of different sizes are given to the tracker. The container size is related to the pixel dimensions of the images. In my case, I believed that all of my images had identical dimensions since they were outputs from a cropping program. However, some of the regions which were being automatically cropped were very close to the image boundary, and so the cropping truncated them, resulting in smaller cropped regions by a few pixels here and there. If you encounter this error, this is the first thing I would check.