search

mal-boisestate / afilament

0 stars 0 forks source link

Run training save log info #40

Open ninanikitina opened 1 year ago

ninanikitina commented 1 year ago

Train and compare 6 different models:

ninanikitina commented 1 year ago

MODEL 1

SET: 42 img of 6 cells by new Zeiss 40x - Alexa488, with Adipo 32 img of 1 initial cell by Zeiss 63x - Alexa488, no Adipo 10 img of 1 cell by Leica 63 x - Alexa488, no Adipo

Weight correction = 200, pos_weight = torch.tensor(200).to(device=device)

INFO: Starting training: Epochs: 200 Batch size: 1 Learning rate: 0.001 Training size: 74 Validation size: 12 Checkpoints: True Device: cuda Images scaling: 1.0

LOG: Mar06_01-00-23_LAPTOP-ISSFJ09KLR_0.001_BS_1_SCALE_1.0

Image

ninanikitina commented 1 year ago

MODEL 2

SET: 42 img of 6 cells by new Zeiss 40x - Alexa488, with Adipo 32 img of 1 initial cell by Zeiss 63x - Alexa488, no Adipo 10 img of 1 cell by Leica 63 x - Alexa488, no Adipo

Weight correction = 20, pos_weight = torch.tensor(20).to(device=device)

INFO: Starting training: Epochs: 200 Batch size: 1 Learning rate: 0.001 Training size: 74 Validation size: 12 Checkpoints: True Device: cuda Images scaling: 1.0

LOG: Mar06_08-40-53_LAPTOP-ISSFJ09KLR_0.001_BS_1_SCALE_1.0

Image

ninanikitina commented 1 year ago

MODEL 3

SET: 42 img of 6 cells by new Zeiss 40x - Alexa488, with Adipo 32 img of 1 initial cell by Zeiss 63x - Alexa488, no Adipo 10 img of 1 cell by Leica 63 x - Alexa488, no Adipo

Weight correction = 2, pos_weight = torch.tensor(2).to(device=device)

INFO: Starting training: Epochs: 200 Batch size: 1 Learning rate: 0.001 Training size: 74 Validation size: 12 Checkpoints: True Device: cuda Images scaling: 1.0

LOG: Mar06_19-37-05_LAPTOP-ISSFJ09KLR_0.001_BS_1_SCALE_1.0

Image

ninanikitina commented 1 year ago

MODEL 4

SET: 32 img of 1 initial cell by Zeiss 63x - Alexa488, no Adipo 10 img of 1 cell by Leica 63 x - Alexa488, no Adipo

Weight correction = 20, pos_weight = torch.tensor(20).to(device=device)

INFO: Starting training: Epochs: 200 Batch size: 1 Learning rate: 0.001 Training size: 74 Validation size: 12 Checkpoints: True Device: cuda Images scaling: 1.0

LOG: Mar07_01-07-32_LAPTOP-ISSFJ09KLR_0.001_BS_1_SCALE_1.0

Image

ninanikitina commented 1 year ago

MODEL 5

SET: 32 img of 1 initial cell by Zeiss 63x - Alexa488, no Adipo

Weight correction = 20, pos_weight = torch.tensor(20).to(device=device)

INFO: Starting training: Epochs: 200 Batch size: 1 Learning rate: 0.001 Training size: 74 Validation size: 12 Checkpoints: True Device: cuda Images scaling: 1.0

LOG: Mar07_05-22-08_LAPTOP-ISSFJ09KLR_0.001_BS_1_SCALE_1.0