search

ChenLiu-1996 / CUTS

[MICCAI 2024] CUTS: A Deep Learning and Topological Framework for Multigranular Unsupervised Medical Image Segmentation
https://arxiv.org/abs/2209.11359
31 stars 2 forks source link

STEGO #7

Closed wwzz-max closed 11 months ago

wwzz-max commented 11 months ago

2 | cluster_probe | ClusterLookup | 700
3 | linear_probe | Conv2d | 710
4 | decoder | Conv2d | 27.3 K 5 | cluster_metrics | UnsupervisedMetrics | 0
6 | linear_metrics | UnsupervisedMetrics | 0
7 | test_cluster_metrics | UnsupervisedMetrics | 0
8 | test_linear_metrics | UnsupervisedMetrics | 0
9 | linear_probe_loss_fn | CrossEntropyLoss | 0
10 | crf_loss_fn | ContrastiveCRFLoss | 0
11 | contrastive_corr_loss_fn | ContrastiveCorrelationLoss | 0

231 K Trainable params 21.7 M Non-trainable params 21.9 M Total params 87.606 Total estimated model params size (MB) Epoch 0: 100%|██████████████████████████████████████████████████████████████████████████████████| 4/4 [00:02<00:00, 1.92it/s, v_num=2]Attempted to log scalar metric test/linear/mIoU:█████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 4.47it/s] nan Attempted to log scalar metric test/linear/Accuracy: nan Attempted to log scalar metric test/cluster/mIoU: nan Attempted to log scalar metric test/cluster/Accuracy: nan Hello, I encountered this problem in STEGO.

ChenLiu-1996 commented 11 months ago

I believe this is the expected behavior. The nan is coming from the fact that no ground truth is provided. In this case, it is learning with self-supervision and produces the corresponding predictions.