potpov/alveolar_canal

Main segmentation project from the paper "Improving Segmentation of the Inferior Alveolar Nerve through Deep Label Propagation" published at CVPR 2022

alveolar_canal's Issues

Wieghts

Hello,

First of all, I found your paper really intresting. Also, thank's for sharing your code.

I was wondering if it is possible to share the weights of your novel segmentation network.

Thank's

Test

Hi,

When testing the model PosPadUNet3D with checkpoints you made available (thank's). I am having a Dice=0.0 and Loss=1. Here is the yaml file that I am using.

title: alveolar_canal_finetuning
project_dir: 'results'
seed: 47

experiment:
  name: Segmentation

data_loader:
  dataset: IAN_Maxillo_dataset\dataset_for_public
  training_set: null
  preprocessing: configs/preprocessing.yaml
  augmentations: configs/augmentations.yaml
  background_suppression: 0
  batch_size: 6
  labels:
    BACKGROUND: 0
    INSIDE: 1
  mean: 0.08435
  num_workers: 4
  patch_shape:
  - 80
  - 80
  - 80
  resize_shape:
  - 168
  - 280
  - 360
  sampler_type: grid
  grid_overlap: 0
  std: 0.17885
  volumes_max: 2100
  volumes_min: 0
  weights:
  - 0.000703
  - 0.999

model:
  name: PosPadUNet3D

loss:
  name: Jaccard

lr_scheduler:
  name: Plateau

optimizer:
  learning_rate: 0.1
  name: SGD

trainer:
  reload: False
  checkpoint: checkpoints\last-seg-pretraining.pth #waht you made available
  do_train: False
  do_test: True
  do_inference: True
  epochs: 100

Do you have any idea why it is not working as expected ?

Regards,
HF

wieghts problems

Hi,
I am having trouble testing the model PosPadUNet3D using a similar conf.yaml that you provided.

title: alveolar_canal_finetuning
project_dir: 'results'
seed: 47

experiment:
  name: Segmentation

data_loader:
  dataset: IAN_Maxillo_dataset\dataset_for_public
  training_set: null
  preprocessing: configs/preprocessing.yaml
  augmentations: configs/augmentations.yaml
  background_suppression: 0
  batch_size: 1
  labels:
    BACKGROUND: 0
    INSIDE: 1
  mean: 0.08435
  num_workers: 4
  patch_shape:
  - 80
  - 80
  - 80
  resize_shape:
  - 168
  - 280
  - 360
  sampler_type: grid
  grid_overlap: 0
  std: 0.17885
  volumes_max: 2100
  volumes_min: 0
  weights:
  - 0.000703
  - 0.999

model:
  name: PosPadUNet3D

loss:
  name: Jaccard

lr_scheduler:
  name: Plateau

optimizer:
  learning_rate: 0.1
  name: Adam

trainer:
  reload: True
  checkpoint: checkpoints\last-seg-pretraining.pth
  do_train: False
  do_test: True
  do_inference: False
  epochs: 100

But still end with theses metrics :

wandb: Run summary:
wandb:     Epoch 29
wandb: Test/Dice 0.0
wandb:  Test/IoU 0.0
wandb: Test/Loss 1.0

PS : Sorry, I created a new issue here. Because I could not create one on AImagelab-zip repository.

Can you please help me.

Regards,
Hamid FSIAN