Adjust history_num and encounter significant performance drop.
Aerotic opened this issue · 4 comments
Greetings,
In train_config.yaml, I've changed the value of history_num to 2, 3, 6. The results show a significant performance drop.
The train and evaluation use the same configuration of history_num.
Is there anything else I need to do to change the value of history_num?
Dear @Aerotic,
I was trying to train the MVDNet after preparing the data and I am encountering some errors. I am not sure how to split the data set for training and validation and where to place them. If you were able to do training, can you please elaborate on steps after preparing the data. Any help is very much appreciated. Thank you.
Hi,
Maybe you should change to a better GPU. 2080Ti works for me
@Aerotic , thanks for the reply. I don't think GPU is a problem for me. The error I am getting is below. I am not sure but I think the issue is with giving the training data paths appropriately. If so, I am not sure how to do it.
File "D:\Implementation\MVDNet\tools\train.py", line 52, in
launch(
File "D:\Implementation\detectron2\detectron2\engine\launch.py", line 62, in launch
main_func(*args)
File "D:\Implementation\MVDNet\tools\train.py", line 44, in main
trainer = Trainer(cfg)
File "D:\Implementation\detectron2\detectron2\engine\defaults.py", line 310, in init
model = self.build_model(cfg)
File "D:\Implementation\detectron2\detectron2\engine\defaults.py", line 452, in build_model
model = build_model(cfg)
File "D:\Implementation\detectron2\detectron2\modeling\meta_arch\build.py", line 21, in build_model
model = META_ARCH_REGISTRY.get(meta_arch)(cfg)
File D:\Implementation\mvdnet\mvdnet\modeling\meta_arch\mvdnet.py", line 21, in init
self.backbone = build_backbone(cfg)
File "D:\Implementation\detectron2\detectron2\modeling\backbone\build.py", line 31, in build_backbone
backbone = BACKBONE_REGISTRY.get(backbone_name)(cfg, input_shape)
File "D:\Implementation\mvdnet\mvdnet\modeling\backbone\mvdnet_backbone.py", line 238, in build_mvdnet_backbone
radar_backbone = VGG(
File "D:\Implementation\mvdnet\mvdnet\modeling\backbone\mvdnet_backbone.py", line 54, in init
...
self.out_channels = out_channels
File "C:/Users/Akash/anaconda3/lib/site-packages/torch/nn/modules/module.py)", line 1317, in setattr
super().setattr(name, value)
AttributeError: can't set attribute
Sorry it has been so long and I forgot the details of experiences about this model