search

ajhamdi / MVTN

pytorch implementation of the ICCV'21 paper "MVTN: Multi-View Transformation Network for 3D Shape Recognition"
98 stars 9 forks source link

Remedy did not work #14

Closed JerkyT closed 1 year ago

JerkyT commented 1 year ago

Hellow，Thanks for your nice work，there is a bug：

i did not change any parameters ，and run ：

CUDA_VISIBLE_DEVICES='2' python run_mvtn.py \
        --data_dir data/ModelNet40/ \
        --run_mode train \
        --mvnetwork mvcnn \
        --nb_views 8 \
        --views_config learned_spherical

output：

...
Evaluation:
        train acc: 82.79 - train Loss: 0.6117
        Val Acc: 76.54 - val Loss: 0.8504
        Current best val acc: 76.90

-----------------------------------
Epoch: [45/100]
        Iter [50/492] Loss: 0.4384
Remedy did not work
ajhamdi commented 1 year ago

hi @JerkyT this problem can happen if the views learned are at singularity 90 degrees which can cause a problem in the rotation matrix used by the renderer. Please try to increase the EXAHSTION_LIMIT=20 in ops.py to EXAHSTION_LIMIT=100 to reduce the chance of this error happening