4 ms·
I’m kind of impressed by how smooth it is, actually. If you watch videos of state-of-the-art object localization NNs, they tend to be EXTREMELY jumpy. These neu
by mitchellgoffpc 7y ago
I’m kind of impressed by how smooth it is, actually. If you watch videos of state-of-the-art object localization NNs, they tend to be EXTREMELY jumpy. These neural nets usually operate on only a single frame at a time, at least in the lower layers, so their predictions tend to jump around a lot from frame to frame (especially when the camera is moving!)
- yarg 7y agoIt worries me in general though, and I think that a higher level of consistently in the results of the parts of the image that don't significantly change between frames seems like a goal worth pursuing. I also think that a deeper understanding of the mechanisms and techniques required to reduce jitter might offer some insights into ways of handling adversarial images.
- yarg 7y agoAnd I do mean insights and not a potential solution. I think it's an issue related to the handling of the spatial discontinuities introduced by the conversion of an effectively continuous reality into a representation consisting of a large number of discrete elements. I think a more in depth understanding of how to navigate organically introduced discontinuities could provide a baseline against which we can look to combat the maliciously introduced ones. It most likely won't be enough - since the problem source is an adaptive and intelligent adversary.