reagan1311 / locate

LOCATE: Localize and Transfer Object Parts for Weakly Supervised Affordance Grounding (CVPR 2023)

locate's Issues

About how to draw the heatmap picture

Thanks for your great work! I got a result KLD=1.252,SIM=0.398,NSS=1.139,But I don't know how to visualize and generate a beautiful heatmap,can you provide some methods?or maybe the python procedures.

About when to make the code public

First of all, thank you very much for your work. You've done an excellent job. I would like to try to reproduce your work recently, so I am curious when you will make your code public.

About Hand_Object Interaction

Thanks for your great work,I'm thinking that is there a way that we can combine affordance with Hand-Object Interactions? but I'm not quite sure about how to combine affordance with Hand-Object Interactions.

Different affordance labels with nearly the same output

Thanks for your great work. When I tested some samples from the AGD20K dataset, I found that despite inputting objects with different affordance labels, the outputs are nearly the same. Here is an example (Left with 'HOLD', Right with 'CUT_WITH'):

Is this a limiation？

Unable to reproduce results in the Unseen setting

Hi,

Thanks for your work! I've tried to run the code and I can reproduce results in the Seen setting. If I run the Unseen checkpoint, I can see

KLD = 1.406
SIM = 0.371
NSS = 1.156

which is correct. However, if I train the Unseen setting from scratch, it seems to have some issue. The command I'm running is

python train.py --data_root ~/AGD20K --divide Unseen

Here is the training log

data_root=/home/ubuntu/AGD20K
save_root=save_models
divide=Unseen
crop_size=224
resize_size=256
num_workers=8
batch_size=16
warm_epoch=0
epochs=15
lr=0.001
momentum=0.9
weight_decay=0.0005
show_step=100
gpu=0
viz=False
test_batch_size=1
test_num_workers=8
num_classes=25                                                                                                                                                                    [120/1265]
exocentric_root=/home/ubuntu/AGD20K/Unseen/trainset/exocentric
egocentric_root=/home/ubuntu/AGD20K/Unseen/trainset/egocentric
test_root=/home/ubuntu/AGD20K/Unseen/testset/egocentric
mask_root=/home/ubuntu/AGD20K/Unseen/testset/GT
save_path=save_models/20230815_055252

train has 13323 images
test has 540 images
Train begining!
LR = [0.001]
epoch: 1/15 + 100/833 | exo_aff_acc: 52.062 | ego_ce: 1.040 | exo_ce: 1.175 | con_loss: 0.000 | loss_cen: 0.048
epoch: 1/15 + 200/833 | exo_aff_acc: 54.812 | ego_ce: 0.800 | exo_ce: 1.036 | con_loss: 0.000 | loss_cen: 0.048
epoch: 1/15 + 300/833 | exo_aff_acc: 55.167 | ego_ce: 0.988 | exo_ce: 1.247 | con_loss: 0.000 | loss_cen: 0.048
epoch: 1/15 + 400/833 | exo_aff_acc: 56.016 | ego_ce: 1.078 | exo_ce: 1.324 | con_loss: 0.000 | loss_cen: 0.048
epoch: 1/15 + 500/833 | exo_aff_acc: 56.888 | ego_ce: 1.145 | exo_ce: 1.246 | con_loss: 0.000 | loss_cen: 0.047
epoch: 1/15 + 600/833 | exo_aff_acc: 57.344 | ego_ce: 1.313 | exo_ce: 1.269 | con_loss: 0.000 | loss_cen: 0.047
epoch: 1/15 + 700/833 | exo_aff_acc: 57.893 | ego_ce: 1.089 | exo_ce: 0.933 | con_loss: 0.000 | loss_cen: 0.047
epoch: 1/15 + 800/833 | exo_aff_acc: 57.938 | ego_ce: 1.223 | exo_ce: 1.019 | con_loss: 0.000 | loss_cen: 0.047
epoch=1 mKLD = 1.957 mSIM = 0.237 mNSS = 0.62 bestKLD = 1000
LR = [0.001]
epoch: 2/15 + 100/833 | exo_aff_acc: 63.312 | ego_ce: 1.465 | exo_ce: 1.068 | con_loss: 0.341 | loss_cen: 0.047
epoch: 2/15 + 200/833 | exo_aff_acc: 61.938 | ego_ce: 0.934 | exo_ce: 0.781 | con_loss: 0.149 | loss_cen: 0.047
epoch: 2/15 + 300/833 | exo_aff_acc: 61.583 | ego_ce: 0.848 | exo_ce: 0.761 | con_loss: 0.201 | loss_cen: 0.047
epoch: 2/15 + 400/833 | exo_aff_acc: 62.391 | ego_ce: 0.966 | exo_ce: 1.204 | con_loss: 0.167 | loss_cen: 0.047
epoch: 2/15 + 500/833 | exo_aff_acc: 62.663 | ego_ce: 0.886 | exo_ce: 0.850 | con_loss: 0.262 | loss_cen: 0.046
epoch: 2/15 + 600/833 | exo_aff_acc: 63.281 | ego_ce: 0.732 | exo_ce: 0.855 | con_loss: 0.316 | loss_cen: 0.046
epoch: 2/15 + 700/833 | exo_aff_acc: 63.482 | ego_ce: 1.028 | exo_ce: 0.996 | con_loss: 0.283 | loss_cen: 0.046
epoch: 2/15 + 800/833 | exo_aff_acc: 63.898 | ego_ce: 1.199 | exo_ce: 0.898 | con_loss: 0.165 | loss_cen: 0.046
epoch=2 mKLD = 1.784 mSIM = 0.28 mNSS = 0.749 bestKLD = 1.957
LR = [0.001]
epoch: 3/15 + 100/833 | exo_aff_acc: 66.562 | ego_ce: 0.789 | exo_ce: 0.997 | con_loss: 0.221 | loss_cen: 0.046
epoch: 3/15 + 200/833 | exo_aff_acc: 66.219 | ego_ce: 0.786 | exo_ce: 0.796 | con_loss: 0.224 | loss_cen: 0.046
epoch: 3/15 + 300/833 | exo_aff_acc: 65.500 | ego_ce: 0.963 | exo_ce: 1.078 | con_loss: 0.311 | loss_cen: 0.046
epoch: 3/15 + 400/833 | exo_aff_acc: 66.031 | ego_ce: 1.081 | exo_ce: 0.923 | con_loss: 0.234 | loss_cen: 0.046
epoch: 3/15 + 500/833 | exo_aff_acc: 66.200 | ego_ce: 0.884 | exo_ce: 0.833 | con_loss: 0.245 | loss_cen: 0.046
epoch: 3/15 + 600/833 | exo_aff_acc: 66.031 | ego_ce: 0.862 | exo_ce: 0.975 | con_loss: 0.171 | loss_cen: 0.046
epoch: 3/15 + 700/833 | exo_aff_acc: 66.223 | ego_ce: 0.806 | exo_ce: 0.848 | con_loss: 0.232 | loss_cen: 0.045
epoch: 3/15 + 800/833 | exo_aff_acc: 66.172 | ego_ce: 0.990 | exo_ce: 0.820 | con_loss: 0.205 | loss_cen: 0.046
epoch=3 mKLD = 1.738 mSIM = 0.29 mNSS = 0.872 bestKLD = 1.784
LR = [0.001]                                                                                                                                                                       [80/1265]
epoch: 4/15 + 100/833 | exo_aff_acc: 67.250 | ego_ce: 1.010 | exo_ce: 0.758 | con_loss: 0.228 | loss_cen: 0.045
epoch: 4/15 + 200/833 | exo_aff_acc: 68.250 | ego_ce: 0.894 | exo_ce: 0.764 | con_loss: 0.230 | loss_cen: 0.045
epoch: 4/15 + 300/833 | exo_aff_acc: 67.875 | ego_ce: 0.631 | exo_ce: 0.692 | con_loss: 0.237 | loss_cen: 0.046
epoch: 4/15 + 400/833 | exo_aff_acc: 67.547 | ego_ce: 0.767 | exo_ce: 0.741 | con_loss: 0.202 | loss_cen: 0.045
epoch: 4/15 + 500/833 | exo_aff_acc: 67.850 | ego_ce: 1.019 | exo_ce: 1.113 | con_loss: 0.163 | loss_cen: 0.045
epoch: 4/15 + 600/833 | exo_aff_acc: 67.760 | ego_ce: 1.034 | exo_ce: 0.679 | con_loss: 0.330 | loss_cen: 0.045
epoch: 4/15 + 700/833 | exo_aff_acc: 68.009 | ego_ce: 0.818 | exo_ce: 0.786 | con_loss: 0.164 | loss_cen: 0.045
epoch: 4/15 + 800/833 | exo_aff_acc: 68.086 | ego_ce: 0.744 | exo_ce: 0.819 | con_loss: 0.211 | loss_cen: 0.045
epoch=4 mKLD = 1.732 mSIM = 0.305 mNSS = 0.878 bestKLD = 1.738
LR = [0.001]
epoch: 5/15 + 100/833 | exo_aff_acc: 69.062 | ego_ce: 0.649 | exo_ce: 0.750 | con_loss: 0.256 | loss_cen: 0.045
epoch: 5/15 + 200/833 | exo_aff_acc: 69.188 | ego_ce: 0.712 | exo_ce: 0.547 | con_loss: 0.234 | loss_cen: 0.045
epoch: 5/15 + 300/833 | exo_aff_acc: 68.208 | ego_ce: 0.861 | exo_ce: 0.905 | con_loss: 0.181 | loss_cen: 0.045
epoch: 5/15 + 400/833 | exo_aff_acc: 68.562 | ego_ce: 1.481 | exo_ce: 1.203 | con_loss: 0.284 | loss_cen: 0.046
epoch: 5/15 + 500/833 | exo_aff_acc: 68.425 | ego_ce: 0.767 | exo_ce: 0.582 | con_loss: 0.114 | loss_cen: 0.045
epoch: 5/15 + 600/833 | exo_aff_acc: 68.740 | ego_ce: 1.034 | exo_ce: 0.757 | con_loss: 0.267 | loss_cen: 0.045
epoch: 5/15 + 700/833 | exo_aff_acc: 68.821 | ego_ce: 0.684 | exo_ce: 0.668 | con_loss: 0.215 | loss_cen: 0.045
epoch: 5/15 + 800/833 | exo_aff_acc: 69.000 | ego_ce: 0.780 | exo_ce: 0.701 | con_loss: 0.206 | loss_cen: 0.045
epoch=5 mKLD = 1.671 mSIM = 0.313 mNSS = 0.949 bestKLD = 1.732
LR = [0.0001]
epoch: 6/15 + 100/833 | exo_aff_acc: 73.562 | ego_ce: 0.743 | exo_ce: 0.527 | con_loss: 0.217 | loss_cen: 0.046
epoch: 6/15 + 200/833 | exo_aff_acc: 71.469 | ego_ce: 0.712 | exo_ce: 0.584 | con_loss: 0.192 | loss_cen: 0.045
epoch: 6/15 + 300/833 | exo_aff_acc: 71.292 | ego_ce: 0.784 | exo_ce: 0.843 | con_loss: 0.265 | loss_cen: 0.046
epoch: 6/15 + 400/833 | exo_aff_acc: 71.453 | ego_ce: 0.635 | exo_ce: 0.456 | con_loss: 0.238 | loss_cen: 0.046
epoch: 6/15 + 500/833 | exo_aff_acc: 71.463 | ego_ce: 0.946 | exo_ce: 0.811 | con_loss: 0.309 | loss_cen: 0.046
epoch: 6/15 + 600/833 | exo_aff_acc: 71.958 | ego_ce: 1.097 | exo_ce: 0.630 | con_loss: 0.269 | loss_cen: 0.045
epoch: 6/15 + 700/833 | exo_aff_acc: 72.107 | ego_ce: 1.089 | exo_ce: 0.615 | con_loss: 0.174 | loss_cen: 0.046
epoch: 6/15 + 800/833 | exo_aff_acc: 72.156 | ego_ce: 0.774 | exo_ce: 0.583 | con_loss: 0.188 | loss_cen: 0.045
epoch=6 mKLD = 1.679 mSIM = 0.315 mNSS = 0.957 bestKLD = 1.671
LR = [0.0001]
epoch: 7/15 + 100/833 | exo_aff_acc: 71.688 | ego_ce: 1.181 | exo_ce: 0.643 | con_loss: 0.173 | loss_cen: 0.046
epoch: 7/15 + 200/833 | exo_aff_acc: 71.719 | ego_ce: 1.058 | exo_ce: 0.465 | con_loss: 0.206 | loss_cen: 0.046
epoch: 7/15 + 300/833 | exo_aff_acc: 72.438 | ego_ce: 0.781 | exo_ce: 0.698 | con_loss: 0.096 | loss_cen: 0.046
epoch: 7/15 + 400/833 | exo_aff_acc: 73.047 | ego_ce: 0.695 | exo_ce: 0.587 | con_loss: 0.151 | loss_cen: 0.045
epoch: 7/15 + 500/833 | exo_aff_acc: 72.838 | ego_ce: 0.941 | exo_ce: 0.704 | con_loss: 0.227 | loss_cen: 0.046
epoch: 7/15 + 600/833 | exo_aff_acc: 72.875 | ego_ce: 0.796 | exo_ce: 0.666 | con_loss: 0.157 | loss_cen: 0.045
epoch: 7/15 + 700/833 | exo_aff_acc: 72.821 | ego_ce: 0.791 | exo_ce: 0.638 | con_loss: 0.345 | loss_cen: 0.046
epoch: 7/15 + 800/833 | exo_aff_acc: 72.883 | ego_ce: 0.708 | exo_ce: 0.557 | con_loss: 0.187 | loss_cen: 0.045
epoch=7 mKLD = 1.694 mSIM = 0.313 mNSS = 0.941 bestKLD = 1.671
LR = [0.0001]                                                                                                                                                                      [40/1265]
epoch: 8/15 + 100/833 | exo_aff_acc: 74.250 | ego_ce: 1.216 | exo_ce: 0.509 | con_loss: 0.149 | loss_cen: 0.046
epoch: 8/15 + 200/833 | exo_aff_acc: 74.156 | ego_ce: 0.890 | exo_ce: 0.818 | con_loss: 0.252 | loss_cen: 0.045
epoch: 8/15 + 300/833 | exo_aff_acc: 73.833 | ego_ce: 0.736 | exo_ce: 0.500 | con_loss: 0.127 | loss_cen: 0.045
epoch: 8/15 + 400/833 | exo_aff_acc: 73.812 | ego_ce: 0.791 | exo_ce: 0.429 | con_loss: 0.220 | loss_cen: 0.046
epoch: 8/15 + 500/833 | exo_aff_acc: 73.537 | ego_ce: 0.909 | exo_ce: 0.669 | con_loss: 0.153 | loss_cen: 0.045
epoch: 8/15 + 600/833 | exo_aff_acc: 72.927 | ego_ce: 1.044 | exo_ce: 0.630 | con_loss: 0.273 | loss_cen: 0.045
epoch: 8/15 + 700/833 | exo_aff_acc: 73.152 | ego_ce: 0.659 | exo_ce: 0.581 | con_loss: 0.231 | loss_cen: 0.046
epoch: 8/15 + 800/833 | exo_aff_acc: 72.781 | ego_ce: 0.822 | exo_ce: 0.493 | con_loss: 0.199 | loss_cen: 0.045
epoch=8 mKLD = 1.667 mSIM = 0.316 mNSS = 0.96 bestKLD = 1.671
LR = [0.0001]
epoch: 9/15 + 100/833 | exo_aff_acc: 71.625 | ego_ce: 0.704 | exo_ce: 0.326 | con_loss: 0.170 | loss_cen: 0.045
epoch: 9/15 + 200/833 | exo_aff_acc: 72.094 | ego_ce: 0.723 | exo_ce: 0.553 | con_loss: 0.234 | loss_cen: 0.046
epoch: 9/15 + 300/833 | exo_aff_acc: 72.938 | ego_ce: 0.668 | exo_ce: 0.635 | con_loss: 0.159 | loss_cen: 0.045
epoch: 9/15 + 400/833 | exo_aff_acc: 73.266 | ego_ce: 1.215 | exo_ce: 0.615 | con_loss: 0.188 | loss_cen: 0.046
epoch: 9/15 + 500/833 | exo_aff_acc: 73.150 | ego_ce: 0.931 | exo_ce: 0.425 | con_loss: 0.264 | loss_cen: 0.046
epoch: 9/15 + 600/833 | exo_aff_acc: 73.021 | ego_ce: 1.154 | exo_ce: 0.714 | con_loss: 0.149 | loss_cen: 0.045
epoch: 9/15 + 700/833 | exo_aff_acc: 73.054 | ego_ce: 1.035 | exo_ce: 0.712 | con_loss: 0.217 | loss_cen: 0.046
epoch: 9/15 + 800/833 | exo_aff_acc: 73.047 | ego_ce: 0.889 | exo_ce: 0.543 | con_loss: 0.114 | loss_cen: 0.045
epoch=9 mKLD = 1.691 mSIM = 0.311 mNSS = 0.933 bestKLD = 1.667
LR = [0.0001]
epoch: 10/15 + 100/833 | exo_aff_acc: 73.875 | ego_ce: 0.842 | exo_ce: 0.662 | con_loss: 0.201 | loss_cen: 0.045
epoch: 10/15 + 200/833 | exo_aff_acc: 73.875 | ego_ce: 0.764 | exo_ce: 0.605 | con_loss: 0.171 | loss_cen: 0.045
epoch: 10/15 + 300/833 | exo_aff_acc: 73.396 | ego_ce: 0.717 | exo_ce: 0.496 | con_loss: 0.209 | loss_cen: 0.045
epoch: 10/15 + 400/833 | exo_aff_acc: 73.984 | ego_ce: 0.711 | exo_ce: 0.506 | con_loss: 0.137 | loss_cen: 0.045
epoch: 10/15 + 500/833 | exo_aff_acc: 73.525 | ego_ce: 0.540 | exo_ce: 0.365 | con_loss: 0.220 | loss_cen: 0.045
epoch: 10/15 + 600/833 | exo_aff_acc: 73.250 | ego_ce: 0.712 | exo_ce: 0.524 | con_loss: 0.180 | loss_cen: 0.045
epoch: 10/15 + 700/833 | exo_aff_acc: 73.339 | ego_ce: 0.752 | exo_ce: 0.516 | con_loss: 0.203 | loss_cen: 0.046
epoch: 10/15 + 800/833 | exo_aff_acc: 73.289 | ego_ce: 1.206 | exo_ce: 0.537 | con_loss: 0.241 | loss_cen: 0.046
epoch=10 mKLD = 1.709 mSIM = 0.308 mNSS = 0.909 bestKLD = 1.667
LR = [1e-05]
epoch: 11/15 + 100/833 | exo_aff_acc: 74.938 | ego_ce: 0.641 | exo_ce: 0.548 | con_loss: 0.227 | loss_cen: 0.046
epoch: 11/15 + 200/833 | exo_aff_acc: 74.344 | ego_ce: 0.746 | exo_ce: 0.522 | con_loss: 0.159 | loss_cen: 0.046
epoch: 11/15 + 300/833 | exo_aff_acc: 73.500 | ego_ce: 0.892 | exo_ce: 0.739 | con_loss: 0.179 | loss_cen: 0.046
epoch: 11/15 + 400/833 | exo_aff_acc: 73.516 | ego_ce: 0.647 | exo_ce: 0.527 | con_loss: 0.192 | loss_cen: 0.045
epoch: 11/15 + 500/833 | exo_aff_acc: 73.588 | ego_ce: 0.943 | exo_ce: 0.481 | con_loss: 0.304 | loss_cen: 0.046
epoch: 11/15 + 600/833 | exo_aff_acc: 73.656 | ego_ce: 0.629 | exo_ce: 0.509 | con_loss: 0.218 | loss_cen: 0.046
epoch: 11/15 + 700/833 | exo_aff_acc: 73.652 | ego_ce: 0.719 | exo_ce: 0.442 | con_loss: 0.195 | loss_cen: 0.045
epoch: 11/15 + 800/833 | exo_aff_acc: 74.016 | ego_ce: 0.743 | exo_ce: 0.557 | con_loss: 0.165 | loss_cen: 0.046
epoch=11 mKLD = 1.722 mSIM = 0.305 mNSS = 0.9 bestKLD = 1.667
LR = [1e-05]
epoch: 12/15 + 100/833 | exo_aff_acc: 73.188 | ego_ce: 0.900 | exo_ce: 0.735 | con_loss: 0.233 | loss_cen: 0.046
epoch: 12/15 + 200/833 | exo_aff_acc: 73.219 | ego_ce: 0.848 | exo_ce: 0.513 | con_loss: 0.245 | loss_cen: 0.046
epoch: 12/15 + 300/833 | exo_aff_acc: 73.604 | ego_ce: 0.789 | exo_ce: 0.719 | con_loss: 0.130 | loss_cen: 0.045
epoch: 12/15 + 400/833 | exo_aff_acc: 73.828 | ego_ce: 0.757 | exo_ce: 0.544 | con_loss: 0.149 | loss_cen: 0.046
epoch: 12/15 + 500/833 | exo_aff_acc: 73.438 | ego_ce: 1.003 | exo_ce: 0.699 | con_loss: 0.184 | loss_cen: 0.046
epoch: 12/15 + 600/833 | exo_aff_acc: 73.292 | ego_ce: 0.760 | exo_ce: 0.635 | con_loss: 0.181 | loss_cen: 0.045
epoch: 12/15 + 700/833 | exo_aff_acc: 73.536 | ego_ce: 0.805 | exo_ce: 0.615 | con_loss: 0.138 | loss_cen: 0.045
epoch: 12/15 + 800/833 | exo_aff_acc: 73.562 | ego_ce: 0.604 | exo_ce: 0.438 | con_loss: 0.224 | loss_cen: 0.046
epoch=12 mKLD = 1.704 mSIM = 0.309 mNSS = 0.916 bestKLD = 1.667
LR = [1e-05]
epoch: 13/15 + 100/833 | exo_aff_acc: 72.688 | ego_ce: 0.785 | exo_ce: 0.586 | con_loss: 0.164 | loss_cen: 0.045
epoch: 13/15 + 200/833 | exo_aff_acc: 73.375 | ego_ce: 0.769 | exo_ce: 0.504 | con_loss: 0.155 | loss_cen: 0.046
epoch: 13/15 + 300/833 | exo_aff_acc: 73.854 | ego_ce: 0.745 | exo_ce: 0.551 | con_loss: 0.167 | loss_cen: 0.046
epoch: 13/15 + 400/833 | exo_aff_acc: 74.031 | ego_ce: 0.732 | exo_ce: 0.524 | con_loss: 0.161 | loss_cen: 0.046
epoch: 13/15 + 500/833 | exo_aff_acc: 74.175 | ego_ce: 0.701 | exo_ce: 0.479 | con_loss: 0.191 | loss_cen: 0.045
epoch: 13/15 + 600/833 | exo_aff_acc: 74.000 | ego_ce: 0.777 | exo_ce: 0.563 | con_loss: 0.130 | loss_cen: 0.045
epoch: 13/15 + 700/833 | exo_aff_acc: 73.768 | ego_ce: 0.640 | exo_ce: 0.475 | con_loss: 0.227 | loss_cen: 0.045
epoch: 13/15 + 800/833 | exo_aff_acc: 73.719 | ego_ce: 0.747 | exo_ce: 0.545 | con_loss: 0.214 | loss_cen: 0.045
epoch=13 mKLD = 1.703 mSIM = 0.309 mNSS = 0.916 bestKLD = 1.667
LR = [1e-05]
epoch: 14/15 + 100/833 | exo_aff_acc: 74.750 | ego_ce: 0.777 | exo_ce: 0.489 | con_loss: 0.206 | loss_cen: 0.046
epoch: 14/15 + 200/833 | exo_aff_acc: 74.375 | ego_ce: 0.657 | exo_ce: 0.605 | con_loss: 0.258 | loss_cen: 0.045
epoch: 14/15 + 300/833 | exo_aff_acc: 73.396 | ego_ce: 0.625 | exo_ce: 0.522 | con_loss: 0.169 | loss_cen: 0.046
epoch: 14/15 + 400/833 | exo_aff_acc: 73.625 | ego_ce: 0.703 | exo_ce: 0.500 | con_loss: 0.300 | loss_cen: 0.046
epoch: 14/15 + 500/833 | exo_aff_acc: 73.638 | ego_ce: 0.614 | exo_ce: 0.440 | con_loss: 0.104 | loss_cen: 0.045
epoch: 14/15 + 600/833 | exo_aff_acc: 73.719 | ego_ce: 0.977 | exo_ce: 0.612 | con_loss: 0.166 | loss_cen: 0.045
epoch: 14/15 + 700/833 | exo_aff_acc: 73.786 | ego_ce: 0.918 | exo_ce: 0.642 | con_loss: 0.161 | loss_cen: 0.045
epoch: 14/15 + 800/833 | exo_aff_acc: 73.414 | ego_ce: 0.860 | exo_ce: 0.583 | con_loss: 0.185 | loss_cen: 0.045
epoch=14 mKLD = 1.702 mSIM = 0.309 mNSS = 0.92 bestKLD = 1.667
LR = [1e-05]
epoch: 15/15 + 100/833 | exo_aff_acc: 74.812 | ego_ce: 0.888 | exo_ce: 0.712 | con_loss: 0.196 | loss_cen: 0.045
epoch: 15/15 + 200/833 | exo_aff_acc: 74.219 | ego_ce: 0.654 | exo_ce: 0.425 | con_loss: 0.223 | loss_cen: 0.046
epoch: 15/15 + 300/833 | exo_aff_acc: 73.417 | ego_ce: 1.016 | exo_ce: 0.627 | con_loss: 0.167 | loss_cen: 0.046
epoch: 15/15 + 400/833 | exo_aff_acc: 73.641 | ego_ce: 1.030 | exo_ce: 0.725 | con_loss: 0.203 | loss_cen: 0.045
epoch: 15/15 + 500/833 | exo_aff_acc: 73.675 | ego_ce: 0.755 | exo_ce: 0.656 | con_loss: 0.160 | loss_cen: 0.045
epoch: 15/15 + 600/833 | exo_aff_acc: 73.688 | ego_ce: 0.711 | exo_ce: 0.617 | con_loss: 0.143 | loss_cen: 0.045
epoch: 15/15 + 700/833 | exo_aff_acc: 73.732 | ego_ce: 0.911 | exo_ce: 0.471 | con_loss: 0.166 | loss_cen: 0.046
epoch: 15/15 + 800/833 | exo_aff_acc: 73.672 | ego_ce: 0.856 | exo_ce: 0.648 | con_loss: 0.154 | loss_cen: 0.045
epoch=15 mKLD = 1.697 mSIM = 0.31 mNSS = 0.922 bestKLD = 1.667

The final performance is significantly worse than the provided checkpoint (KLD 1.667 vs 1.406). Do you see anything suspcious?