ClimateGArt's avatar

ClimateGArt

A display of our dramatic yet mesmerizing failures training ClimateGAN

Do not underestimate the output of a faulty model.

This website gathers creative yet unvoluntary visual findings along our two years of research creating ClimateGAN.

17_09
2880pxassemblee
wildfire_failure
zebromnigan

A storm of pixels. Can you survive Hurricane Colora?

These images result from a faulty post-processing, rescaling [0; 1] pixels as if they were [0; 255].

14_54_wildfire
2a4pv_wildfire
ab_231_wildfire
daewgf_wildfire
holland_building

Virtual Pointillism.

A segmentation model can be creative too you know? Especially if you train it with the wrong loss.

real-pointillisme-2
real-pointillisme-3
real-pointillisme-4
sim-pointillisme-0
sim-pointillisme-2
sim-pointillisme-3

Saturation is all you need.

GANs are unstable beasts. They tend to collapse, and God knows where they land when you forget the discriminator.

003
013
014

Stumbling upon your own failures can lift your spirits, when you are open to surprises.

A random collection of failures and accidents in our quest for depth maps.

purplehaze
futuristicollage
zebradepth

Logging is a timber's burden

There's a dark forest of traps ahead: axing your way through model comparison and multi-modal visual outputs plotting is a journey in itself.

007582_3_step_3
007647_1_step_9
jr_250_step_78
lisbon1_step_86
polk_street_concrete_cottages_step_98
driveway-3088488_960_720_step_133
real-estate-475875_960_720_step_164
train_r_s_12_step_12
train_s_s_12_step_12
val_r_s_12_step_12
val_s_s_12_step_12

GANs breathe too. Their heartbeats soothe the soul.

Forwarding a series of noise images through a GauGAN non-trained (just initialized) model. When the center pixel is artificially boosted, the model naturally focuses on that part of the noise, creating this narrower and narrower output circle.