How to use prototypes.pth that only contains a single class?

[kaninaba94]

I want to train a few shot object detector to recognize only a single class, which I built prototypes for using this blueprint: https://github.com/mlzxy/devit/blob/main/demo/build_prototypes.ipynb

Now when I run python demo/demo.py with some other images of the same class as input, I get an error, presumably because I only have a single class in my prototypes:

File "demo/demo.py", line 272, in <module>
    fire.Fire(main)
  File "/home/appuser/.local/lib/python3.8/site-packages/fire/core.py", line 141, in Fire
    component_trace = _Fire(component, args, parsed_flag_args, context, name)
  File "/home/appuser/.local/lib/python3.8/site-packages/fire/core.py", line 475, in _Fire
    component, remaining_args = _CallAndUpdateTrace(
  File "/home/appuser/.local/lib/python3.8/site-packages/fire/core.py", line 691, in _CallAndUpdateTrace
    component = fn(*varargs, **kwargs)
  File "demo/demo.py", line 221, in main
    output = model(batched_inputs)[0]
  File "/home/appuser/.local/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1194, in _call_impl
    return forward_call(*input, **kwargs)
  File "/home/appuser/devit/demo/../detectron2/modeling/meta_arch/devit.py", line 1168, in forward
    other_classes = interpolate(other_classes, self.T, mode='linear') # (Nxclasses) x spatial x T
  File "/home/appuser/devit/demo/../detectron2/modeling/meta_arch/devit.py", line 86, in interpolate
    return F.interpolate(seq, T, mode=mode) 
  File "/home/appuser/.local/lib/python3.8/site-packages/torch/nn/functional.py", line 3945, in interpolate
    return torch._C._nn.upsample_linear1d(input, output_size, align_corners, scale_factors)
RuntimeError: Input and output sizes should be greater than 0, but got input (W: 0) and output (W: 128)

Any hunch as to how I can solve this? Might the issue be that I did not explicitly train few shot object detector using https://github.com/mlzxy/devit/blob/main/tools/train_net.py ?

[Rivoks]

See here: https://github.com/mlzxy/devit/issues/8#issuecomment-1754145097

[renshouyu]

I want to train a few shot object detector to recognize only a single class, which I built prototypes for using this blueprint: https://github.com/mlzxy/devit/blob/main/demo/build_prototypes.ipynb

Now when I run python demo/demo.py with some other images of the same class as input, I get an error, presumably because I only have a single class in my prototypes:

File "demo/demo.py", line 272, in <module>
    fire.Fire(main)
  File "/home/appuser/.local/lib/python3.8/site-packages/fire/core.py", line 141, in Fire
    component_trace = _Fire(component, args, parsed_flag_args, context, name)
  File "/home/appuser/.local/lib/python3.8/site-packages/fire/core.py", line 475, in _Fire
    component, remaining_args = _CallAndUpdateTrace(
  File "/home/appuser/.local/lib/python3.8/site-packages/fire/core.py", line 691, in _CallAndUpdateTrace
    component = fn(*varargs, **kwargs)
  File "demo/demo.py", line 221, in main
    output = model(batched_inputs)[0]
  File "/home/appuser/.local/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1194, in _call_impl
    return forward_call(*input, **kwargs)
  File "/home/appuser/devit/demo/../detectron2/modeling/meta_arch/devit.py", line 1168, in forward
    other_classes = interpolate(other_classes, self.T, mode='linear') # (Nxclasses) x spatial x T
  File "/home/appuser/devit/demo/../detectron2/modeling/meta_arch/devit.py", line 86, in interpolate
    return F.interpolate(seq, T, mode=mode) 
  File "/home/appuser/.local/lib/python3.8/site-packages/torch/nn/functional.py", line 3945, in interpolate
    return torch._C._nn.upsample_linear1d(input, output_size, align_corners, scale_factors)
RuntimeError: Input and output sizes should be greater than 0, but got input (W: 0) and output (W: 128)

Any hunch as to how I can solve this? Might the issue be that I did not explicitly train few shot object detector using https://github.com/mlzxy/devit/blob/main/tools/train_net.py ?

Hello, have you solved this problem?

[pooya-github]

I want to train a few shot object detector to recognize only a single class, which I built prototypes for using this blueprint: https://github.com/mlzxy/devit/blob/main/demo/build_prototypes.ipynb Now when I run python demo/demo.py with some other images of the same class as input, I get an error, presumably because I only have a single class in my prototypes:

File "demo/demo.py", line 272, in <module>
    fire.Fire(main)
  File "/home/appuser/.local/lib/python3.8/site-packages/fire/core.py", line 141, in Fire
    component_trace = _Fire(component, args, parsed_flag_args, context, name)
  File "/home/appuser/.local/lib/python3.8/site-packages/fire/core.py", line 475, in _Fire
    component, remaining_args = _CallAndUpdateTrace(
  File "/home/appuser/.local/lib/python3.8/site-packages/fire/core.py", line 691, in _CallAndUpdateTrace
    component = fn(*varargs, **kwargs)
  File "demo/demo.py", line 221, in main
    output = model(batched_inputs)[0]
  File "/home/appuser/.local/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1194, in _call_impl
    return forward_call(*input, **kwargs)
  File "/home/appuser/devit/demo/../detectron2/modeling/meta_arch/devit.py", line 1168, in forward
    other_classes = interpolate(other_classes, self.T, mode='linear') # (Nxclasses) x spatial x T
  File "/home/appuser/devit/demo/../detectron2/modeling/meta_arch/devit.py", line 86, in interpolate
    return F.interpolate(seq, T, mode=mode) 
  File "/home/appuser/.local/lib/python3.8/site-packages/torch/nn/functional.py", line 3945, in interpolate
    return torch._C._nn.upsample_linear1d(input, output_size, align_corners, scale_factors)
RuntimeError: Input and output sizes should be greater than 0, but got input (W: 0) and output (W: 128)

Any hunch as to how I can solve this? Might the issue be that I did not explicitly train few shot object detector using https://github.com/mlzxy/devit/blob/main/tools/train_net.py ?

Hello, have you solved this problem?

Hi,

In the file demo.py modify the code in this part:

if category_space is not None: category_space = torch.load(category_space) model.label_names = category_space['label_names'] model.test_class_weight = category_space['prototypes'].to(device)

to:

if category_space is not None: category_space = torch.load(category_space) if len(category_space["label_names"]) < 2: category_space["label_names"].append("blank") real_prototypes = category_space["prototypes"] blank_prototypes = torch.zeros(1, 1024) category_space["prototypes"] = torch.cat((real_prototypes, blank_prototypes), 0) model.label_names = category_space["label_names"] model.test_class_weight = category_space["prototypes"].to(device)