Encoding presets

As with conventional codecs, there is a trade-off between Cool-chic encoding time and compression performance. The encoding settings of Cool-chic are changed using the following arguments:

• --recipe

• --start_lr

• --n_itr

• --n_train_loops

Recipes

Cool-chic encoding works with tweakable recipes i.e. different training parameters. Currently available recipes are:

• c3 Inspired by C3: High-performance and low-complexity neural compression from a single image or video, Kim et al * Composed of two main phases: 1. additive noise model for the quantization 2. Actual quantization

• debug Extremely fast preset with very bad performance only for debugging purposes.

All recipes feature a decreasing learning rate starting from --start_lr.

The number of iterations in the first (and longest) phase of the c3``recipe is set using ``--n_itr.

In order to circumvent some training irregularities, it is possible to perform several independent encoding, keeping only the best one. We call that a training loop. The number of training loops is set by --n_train_loops.

Reasonable arguments

The following encoding parameters gives most of the performance while being reasonably fast.

(venv) python src/encode.py --recipe=c3 --start_lr=1e-2 --n_itr=10000 --n_train_loops=2

Tip

Obtaining the absolute best compression performance from Cool-chic might require significantly longer training than suggested above