search

ByungKwanLee / Causal-Unsupervised-Segmentation

Official PyTorch Implementation code for realizing the technical part of Causal Unsupervised Semantic sEgmentation (CAUSE) to improve performance of unsupervised semantic segmentation. (Under Review)
8 stars 1 forks source link

DINO pretrained model #9

Open Taocunguo opened 4 months ago

Taocunguo commented 4 months ago

Thank you for your outstanding work. I would like to ask whether the DINO pretrained model used is the "backbone only" of the page you gave or "full ckpt". If it is the "backbone only", Runtime message _IncompatibleKeys(missing_keys=['head.weight', 'head.bias'], unexpected_keys=[])

ByungKwanLee commented 4 months ago

We used backbone only!

Taocunguo commented 4 months ago

We used backbone only!

Thank you for your reply. I have another question. When I run test_tr.py, it freezes on the second TEST. Can you tell me why and how to fix it.

------------------Configurations------------------
NAME_TAG: CAUSE-TR
data_dir: /home/tcg/Paper/Causal-Unsupervised-Segmentation/Data
dataset: cocostuff27
port: 12355
ckpt: checkpoint/dino_vit_base_8.pth
distributed: False
load_segment: True
load_cluster: True
train_resolution: 320
test_resolution: 320
batch_size: 16
num_workers: 14
gpu: 0,1,2,3
num_codebook: 2048
reduced_dim: 90
projection_dim: 2048
dim: 768
num_queries: 1600
-------------------------------------------------
_IncompatibleKeys(missing_keys=['head.weight', 'head.bias'], unexpected_keys=[])
[Segment] CAUSE/cocostuff27/dino_vit_base_8/2048/segment_tr.pth loaded
[Cluster] CAUSE/cocostuff27/dino_vit_base_8/2048/cluster_tr.pth loaded
Modularity CAUSE/cocostuff27/modularity/dino_vit_base_8/2048/modular.npy loaded
# of Parameters: 9.84(M)
[TEST] Acc (Linear): 78.1% | [mIoU]: 39.3, [mAP]: 52.2, [Acc]: 72.7, : 100%|█| 136/136 [00:21<00:00,  6.21it
  0%|                                                                                  | 0/136 [00:00<?, ?it
ByungKwanLee commented 4 months ago

Would you more explain the detail of "When I run test_tr.py, it freezes on the second TEST." I didn't catch

Taocunguo commented 4 months ago

Would you more explain the detail of "When I run test_tr.py, it freezes on the second TEST." I didn't catch

When I run test_tr.py, two lines appear. The first line seems to complete the test, but the next line has no result, and the program seems to stop there

[TEST] Acc (Linear): 78.1% | [mIoU]: 39.3, [mAP]: 52.2, [Acc]: 72.7, : 100%|█| 136/136 [00:21<00:00,  6.21it
  0%|                                                                                  | 0/136 [00:00<?, ?it
ByungKwanLee commented 4 months ago

The second lines are representing the CRF version results. I think it is a problem of pydensecrf library!

Taocunguo commented 4 months ago

pydensecrf

oh no,can you tell me your pydensecrf's version,thank you!