The point here on viewing filters as tools for detection underscores the power of convolutional neural networks in modern image processing. By stacking these filters and training them using ML models, we can build powerful feature detectors that can quite accurately perform tasks such as facial detection.
GarciaEricS
I'm still working on 189 HW6 (sad), and this slide really made me understand what convolutions actually are and why they're so important to images in particular. It lets us pick out patterns that exist for particular regions in a way that is very general. By combining these kernels, we can pick out so many interesting patterns that would be harder to detect by looking at pixels individually. This stuff is really cool and makes me excited to take computer vision in the future.
The point here on viewing filters as tools for detection underscores the power of convolutional neural networks in modern image processing. By stacking these filters and training them using ML models, we can build powerful feature detectors that can quite accurately perform tasks such as facial detection.
I'm still working on 189 HW6 (sad), and this slide really made me understand what convolutions actually are and why they're so important to images in particular. It lets us pick out patterns that exist for particular regions in a way that is very general. By combining these kernels, we can pick out so many interesting patterns that would be harder to detect by looking at pixels individually. This stuff is really cool and makes me excited to take computer vision in the future.