search

jiwoncpark / baobab

Generates the training data for hierarchical inference of strongly-lensed systems with Bayesian neural networks
MIT License
11 stars 2 forks source link

Enable "no noise" option #41

Closed jiwoncpark closed 4 years ago

jiwoncpark commented 4 years ago

In generating the training set, we sometimes want to NOT add any noise so we can add them online during training. Now you can do this by setting, in your user config file, cfg.image.add_noise = False. Note that this is different from setting cfg.observation.background_noise = 0.0 because background noise only includes sky and readout noise and not CCD noise -- so Poisson noise will still be added. Current hack is another if statement in sim_utils/image_utils.py -- I'll wrap this in a class so things are set upon instantiation. @swagnercarena

jiwoncpark commented 4 years ago

We decided to hardcode zero noise. Closing...