gordicaleksa / pytorch-deepdream Goto Github PK

Hi, when I try to run the following command I am returned in error:
python3 deepdream.py --input cl_00001.png --img_width 1920 --pyramid_size 2 --model_name RESNET50 --pretrained_weights PLACES_365
Dreaming started!
Invalid layer names ['relu4_3'].
Available layers for model RESNET50 are ['layer1', 'layer2', 'layer3', 'layer4'].
Traceback (most recent call last):
File "/home/reilly/Documents/Github/pytorch-deepdream/deepdream.py", line 236, in
dump_path = utils.save_and_maybe_display_image(config, img)
File "/home/reilly/Documents/Github/pytorch-deepdream/utils/utils.py", line 87, in save_and_maybe_display_image
assert isinstance(dump_img, np.ndarray), f'Expected numpy array got {type(dump_img)}.'
AssertionError: Expected numpy array got <class 'NoneType'>.

Again, any insight into how to resolve this would be greatly appreciated, thank you so much!

I'm making a pull request to fix this.

My GPU is only 30%~50% utilized when running this project, it's too slow!

Hi, when I try to run section 21 of the jupyter notebook I am returned the follow error:
Downloading resnet50_places365.pth.tar from http://places2.csail.mit.edu/models_places365/resnet50_places365.pth.tar it may take some time.

NameError Traceback (most recent call last)
/tmp/ipykernel_212303/229011906.py in
10 config['spatial_shift_size'] = 40
11
---> 12 img = deep_dream_static_image(config)
13 dump_path = save_and_maybe_display_image(config, img)
14 print(f'Saved DeepDream static image to: {os.path.relpath(dump_path)}\n')

/tmp/ipykernel_212303/1103483755.py in deep_dream_static_image(config, img)
1 def deep_dream_static_image(config, img=None):
----> 2 model = fetch_and_prepare_model(config['model_name'], config['pretrained_weights'])
3
4 try:
5 layer_ids_to_use = [model.layer_names.index(layer_name) for layer_name in config['layers_to_use']]

/tmp/ipykernel_212303/3540351205.py in fetch_and_prepare_model(model_type, pretrained_weights)
128 elif model_type == SupportedModels.RESNET50.name:
129 # We'll define the ResNet50 later
--> 130 model = ResNet50(pretrained_weights, requires_grad=False, show_progress=True).to(DEVICE)
131 else:
132 raise Exception('Model not yet supported.')

/tmp/ipykernel_212303/2427502766.py in init(self, pretrained_weights, requires_grad, show_progress)
17 binary_url = r'http://places2.csail.mit.edu/models_places365/resnet50_places365.pth.tar'
18 print(f'Downloading {binary_name} from {binary_url} it may take some time.')
---> 19 download_url_to_file(binary_url, resnet50_places365_binary_path)
20 print('Done downloading.')
21 state_dict = torch.load(resnet50_places365_binary_path)['state_dict']

NameError: name 'download_url_to_file' is not defined

Any idea how to fix this? Thanks!

Thanks for sharing your code. My goal is to use GoogleNet on face data. You mentioned that the result of the model depends on the dataset. Does this mean if I want to see components of human face in the deep dream images I need to train the model from scratch on human face. I tried to fine-tune the model although it gave me good accuracy but it didn't give me dream-like images and if it gave me any dream like image it contained human face components. Do you have any idea or suggestion how to approach this?

!python deepdream.py --input pink.jpg --img_width 2000 --num_gradient_ascent_iterations 10 --pyramid_size 4 --pyramid_ratio 1.2 --smoothing_coefficient 0.5 --model SupportedModels=0 --pretrained_weights IMAGNET

usage: deepdream.py [-h] [--input INPUT] [--img_width IMG_WIDTH]
[--model {SupportedModels.VGG16,SupportedModels.VGG16_EXPERIMENTAL,SupportedModels.GOOGLENET,SupportedModels.RESNET50,SupportedModels.ALEXNET}]
[--pretrained_weights {SupportedPretrainedWeights.IMAGENET,SupportedPretrainedWeights.PLACES_365}]
[--layers_to_use LAYERS_TO_USE]
[--pyramid_size PYRAMID_SIZE]
[--pyramid_ratio PYRAMID_RATIO]
[--num_gradient_ascent_iterations NUM_GRADIENT_ASCENT_ITERATIONS]
[--lr LR] [--is_video IS_VIDEO]
[--video_length VIDEO_LENGTH]
[--frame_transform {SupportedTransforms.ZOOM,SupportedTransforms.ZOOM_ROTATE,SupportedTransforms.TRANSLATE}]
[--blend BLEND] [--should_display SHOULD_DISPLAY]
[--spatial_shift_size SPATIAL_SHIFT_SIZE]
[--smoothing_coefficient SMOOTHING_COEFFICIENT]
[--use_noise USE_NOISE]
deepdream.py: error: argument --model: invalid choice: 'SupportedModels=0' (choose from <SupportedModels.VGG16: 0>, <SupportedModels.VGG16_EXPERIMENTAL: 1>, <SupportedModels.GOOGLENET: 2>, <SupportedModels.RESNET50: 3>, <SupportedModels.ALEXNET: 4>)

what am i doing wrong?? 😞😪

I've attempted to look through this repository, but I have not found any support for this.

I trained googlenet model based on my own image set according to this tutorial: https://pythonawesome.com/quickly-and-easily-create-train-a-custom-deepdream-model/.

Is this currently supported?

Hello! Firstly, let me just thank you for this great repo. Its the clearest practical explanation of Deep Dream I've found online!

The issue:
When I try to run the following command (on a CPU-only machine)

python deepdream.py model_name RESNET50 --pretrained_weights PLACES_365 --layers_to_use layer3

I get the following error

Dreaming started!
Traceback (most recent call last):
  File "C:/Users/Asus/Desktop/My stuff/DeepDream NFT/pytorch-deepdream/deepdream.py", line 235, in <module>
    img = deep_dream_static_image(config, img=None)  # img=None -> will be loaded inside of deep_dream_static_image
  File "C:/Users/Asus/Desktop/My stuff/DeepDream NFT/pytorch-deepdream/deepdream.py", line 69, in deep_dream_static_image
    model = utils.fetch_and_prepare_model(config['model_name'], config['pretrained_weights'], DEVICE)
  File "C:\Users\Asus\Desktop\My stuff\DeepDream NFT\pytorch-deepdream\utils\utils.py", line 132, in fetch_and_prepare_model
    model = ResNet50(pretrained_weights, requires_grad=False, show_progress=True).to(device)
  File "C:\Users\Asus\Desktop\My stuff\DeepDream NFT\pytorch-deepdream\models\definitions\resnets.py", line 28, in __init__
    state_dict = torch.load(resnet50_places365_binary_path)['state_dict']
  File "C:\Users\Asus\Desktop\My stuff\DeepDream NFT\DeepDream_venv\lib\site-packages\torch\serialization.py", line 713, in load
    return _legacy_load(opened_file, map_location, pickle_module, **pickle_load_args)
  File "C:\Users\Asus\Desktop\My stuff\DeepDream NFT\DeepDream_venv\lib\site-packages\torch\serialization.py", line 930, in _legacy_load
    result = unpickler.load()
  File "C:\Users\Asus\Desktop\My stuff\DeepDream NFT\DeepDream_venv\lib\site-packages\torch\serialization.py", line 876, in persistent_load
    wrap_storage=restore_location(obj, location),
  File "C:\Users\Asus\Desktop\My stuff\DeepDream NFT\DeepDream_venv\lib\site-packages\torch\serialization.py", line 176, in default_restore_location
    result = fn(storage, location)
  File "C:\Users\Asus\Desktop\My stuff\DeepDream NFT\DeepDream_venv\lib\site-packages\torch\serialization.py", line 152, in _cuda_deserialize
    device = validate_cuda_device(location)
  File "C:\Users\Asus\Desktop\My stuff\DeepDream NFT\DeepDream_venv\lib\site-packages\torch\serialization.py", line 136, in validate_cuda_device
    raise RuntimeError('Attempting to deserialize object on a CUDA '
RuntimeError: Attempting to deserialize object on a CUDA device but torch.cuda.is_available() is False. If you are running on a CPU-only machine, please use torch.load with map_location=torch.device('cpu') to map your storages to the CPU.

After looking into it, I think the issue may be that the saved model contains GPU tensors. This would make torch.load load those tensors to GPU by default, which causes the CUDA error.

Adding the argument map_location='cpu' to torch.load fixes the issue. I can create a pull request with this fix if you agree with it.