I want to train segfix on my custom dataset. The dataset contains 10 classes and the image size is 2048X2048. While going through the script, I found this "export dt_num_classes=8" in run_h_48_d_4_segfix.sh. Do I need to change this to 10?
Also there's this line "assert num_classes in (4, 8, 16, 32,)" in offset_helper.py. What parameters do I need to change and where?
I tried to change some of these parameters to 10 but I recieved the error
Requirement already satisfied: yacs in /home/workspace/projects/env_aseg/lib/python3.6/site-packages (0.1.8)
Requirement already satisfied: PyYAML in /home/workspace/projects/env_aseg/lib/python3.6/site-packages (from yacs) (5.4.1)
Requirement already satisfied: torchcontrib in /home/workspace/projects/env_aseg/lib/python3.6/site-packages (0.0.2)
Requirement already satisfied: pydensecrf in /home/workspace/projects/env_aseg/lib/python3.6/site-packages (1.0rc3)
dt_max_distance: 5
dt_num_classes: 8
offset_dir: offset_gt/dt_offset
Logging to ./log/cityscapes/segfix_hrnet_hrnet48_segfix_loss_1.log
2021-07-17 17:54:46,324 INFO [offset_helper.py, 55] engery/max-distance: 5 engery/min-distance: 0
2021-07-17 17:54:46,324 INFO [offset_helper.py, 62] direction/num_classes: 8 scale: 1
2021-07-17 17:54:46,324 INFO [offset_helper.py, 67] c4 align axis: False
2021-07-17 17:54:46,334 INFO [module_runner.py, 44] BN Type is inplace_abn.
2021-07-17 17:54:46,334 INFO [init.py, 17] Using evaluator: StandardEvaluator
Traceback (most recent call last):
File "main.py", line 214, in
model = Trainer(configer)
File "/home/workspace/projects/segfix_falcon/openseg.pytorch/segmentor/trainer.py", line 70, in init
self._init_model()
File "/home/workspace/projects/segfix_falcon/openseg.pytorch/segmentor/trainer.py", line 73, in _init_model
self.seg_net = self.model_manager.semantic_segmentor()
File "/home/workspace/projects/segfix_falcon/openseg.pytorch/lib/models/model_manager.py", line 86, in semantic_segmentor
model = SEG_MODEL_DICTmodel_name
File "/home/workspace/projects/segfix_falcon/openseg.pytorch/lib/models/nets/segfix.py", line 29, in init
self.backbone = BackboneSelector(configer).get_backbone()
File "/home/workspace/projects/segfix_falcon/openseg.pytorch/lib/models/backbones/backbone_selector.py", line 34, in get_backbone
model = HRNetBackbone(self.configer)(**params)
File "/home/workspace/projects/segfix_falcon/openseg.pytorch/lib/models/backbones/hrnet/hrnet_backbone.py", line 769, in call
bn_momentum=0.1)
File "/home/workspace/projects/segfix_falcon/openseg.pytorch/lib/models/backbones/hrnet/hrnet_backbone.py", line 313, in init
self.bn1 = ModuleHelper.BatchNorm2d(bn_type=bn_type)(64, momentum=bn_momentum)
TypeError: 'NoneType' object is not callable
I want to train segfix on my custom dataset. The dataset contains 10 classes and the image size is 2048X2048. While going through the script, I found this "export dt_num_classes=8" in run_h_48_d_4_segfix.sh. Do I need to change this to 10? Also there's this line "assert num_classes in (4, 8, 16, 32,)" in offset_helper.py. What parameters do I need to change and where?
I tried to change some of these parameters to 10 but I recieved the error
Requirement already satisfied: yacs in /home/workspace/projects/env_aseg/lib/python3.6/site-packages (0.1.8) Requirement already satisfied: PyYAML in /home/workspace/projects/env_aseg/lib/python3.6/site-packages (from yacs) (5.4.1) Requirement already satisfied: torchcontrib in /home/workspace/projects/env_aseg/lib/python3.6/site-packages (0.0.2) Requirement already satisfied: pydensecrf in /home/workspace/projects/env_aseg/lib/python3.6/site-packages (1.0rc3) dt_max_distance: 5 dt_num_classes: 8 offset_dir: offset_gt/dt_offset Logging to ./log/cityscapes/segfix_hrnet_hrnet48_segfix_loss_1.log 2021-07-17 17:54:46,324 INFO [offset_helper.py, 55] engery/max-distance: 5 engery/min-distance: 0 2021-07-17 17:54:46,324 INFO [offset_helper.py, 62] direction/num_classes: 8 scale: 1 2021-07-17 17:54:46,324 INFO [offset_helper.py, 67] c4 align axis: False 2021-07-17 17:54:46,334 INFO [module_runner.py, 44] BN Type is inplace_abn. 2021-07-17 17:54:46,334 INFO [init.py, 17] Using evaluator: StandardEvaluator Traceback (most recent call last): File "main.py", line 214, in
model = Trainer(configer)
File "/home/workspace/projects/segfix_falcon/openseg.pytorch/segmentor/trainer.py", line 70, in init
self._init_model()
File "/home/workspace/projects/segfix_falcon/openseg.pytorch/segmentor/trainer.py", line 73, in _init_model
self.seg_net = self.model_manager.semantic_segmentor()
File "/home/workspace/projects/segfix_falcon/openseg.pytorch/lib/models/model_manager.py", line 86, in semantic_segmentor
model = SEG_MODEL_DICTmodel_name
File "/home/workspace/projects/segfix_falcon/openseg.pytorch/lib/models/nets/segfix.py", line 29, in init
self.backbone = BackboneSelector(configer).get_backbone()
File "/home/workspace/projects/segfix_falcon/openseg.pytorch/lib/models/backbones/backbone_selector.py", line 34, in get_backbone
model = HRNetBackbone(self.configer)(**params)
File "/home/workspace/projects/segfix_falcon/openseg.pytorch/lib/models/backbones/hrnet/hrnet_backbone.py", line 769, in call
bn_momentum=0.1)
File "/home/workspace/projects/segfix_falcon/openseg.pytorch/lib/models/backbones/hrnet/hrnet_backbone.py", line 313, in init
self.bn1 = ModuleHelper.BatchNorm2d(bn_type=bn_type)(64, momentum=bn_momentum)
TypeError: 'NoneType' object is not callable