Kornia VS Dlib

👉 https://github.com/kornia/kornia/releases/tag/v0.6.6

Kornia: Differential Computer Vision\ (3 comments)

Kornia - Kornia is a differentiable computer vision library for PyTorch. It consists of a set of routines and differentiable modules to solve generic computer vision problems. At its core, the package uses PyTorch as its main backend both for efficiency and to take advantage of the reverse-mode auto-differentiation to define and compute the gradient of complex functions.

Between "not being sure yet" about GPU operations in pre-processing and choosing high-end CPUs, I think you are overthinking the wrong alternative. Besides DALI, check whether you are using codecs besides nvidia/torchvision-supported jpeg and png, and if other GPU CV libraries meet your needs: torchvision kornia

Use kornia.augmentation where this problem is solved doing the augmentations in batch outside the dataloader. https://github.com/kornia/kornia

dlib

The apparent gender estimates from photos are using dlib, and I really ought to get what I'm doing cleaned up in such a way that other people can use it easily.

I really like dlib's code https://github.com/davisking/dlib

Additionally, C++ may be used for extremely high levels of optimization even for cloud-based ML. Dlib and Kaldi are C++ libraries used as dependencies in Python codebases for computer vision and audio processing, for example. So if your application requires you to customize any functions similar to those libraries, then you'll need C++ knowhow.

If you know C++, you don't need anything else. Go and learn APIs for C++ libraries. If you're into DSP, why not study Dlib?.

The data is mostly in this spreadsheet. The apparently facial gender estimates are made with Dlib. The mental health assessments are from Beck's Depression Inventory and the Snaith-Hamilton Pleasure Scale. The graph is made with gnuplot.

The plugin uses dlib library with a very fast HOG detector for both face recognition and detector following the relative examples.

The dlib facial recognition model thinks that I am now a distance of about 0.3 from where I started, which is far enough to start getting many false positive matches, but still within the design intent that different pictures of the same individual will be within 0.6 of each other.

Dlib's face recognition module thinks that I am about 0.25 units away from where I started; its design intent is that distinct individuals will be 0.6 or more apart, although in practice other people start showing up around 0.3.