stinkbug experiments

[AlexEMG]

• create 3 splits -> @AlexEMG
• create @biol-jsh
• variant @Kadzon
• variant @sabrinabenas
• variant @sfmig
• variant @neslihanedes #2
• variant @vchiang001

[AlexEMG]

@AlexEMG Train:

• main_train_baseline.py
  gputouse=2

[biol-jsh]

@AlexEMG Train:

• gauss_noise_train.py
  here gputouse=0

[vchiang001]

@AlexEMG

• removed rotation

gpu = 1

[sabrinabenas]

@AlexEMG

• removed covering and convolution

gpu = 3

[Kadzon]

@AlexEMG

• changed scale to no scale

gpu=3

[AlexEMG]

nohup python3 removedCovering_addConvolution.py > log_sabrina.txt &
nohup python3 train_without_scale.py > log_nirel.txt &

[neslihanedes]

@AlexEMG

• 1st model- graysale: True
• 2nd model- motion_blur: False
• 3rd model- mirror: True
  I can't commit my notebook now from the server (no commit rights). Here will be updated on Monday 🐦

[AlexEMG]

Finalize the runs:

• run & evaluate 3 splits @biol-jsh
• run & evaluate variant @Kadzon
• run & evaluate variant @sabrinabenas
• run & evaluate variant @sfmig
• run & evaluate variant @vchiang001
• run & evaluate variant @neslihanedes

[biol-jsh]

The single animal version of these are all on the way. We are currently training all augmentation variants, 3 shuffles, 300,000 iterations.