vkhoi / kth-action-recognition Goto Github PK

I have installed the latest OpenCV version-> 4.6.0.66 and Pytorch version->1.12.1.
I am getting the following error.

[ WARN:[email protected]] global D:\a\opencv-python\opencv-python\opencv\modules\imgcodecs\src\loadsave.cpp (239) cv::findDecoder imread_('..\dataset\walking\person25_walking_d4_uncomp.avi'): can't open/read file: check file path/integrity

Kindly find a screenshot of the error.

Please help me to resolve this error.

I can't make the dataset. The following exception is raised. Though I have updated imageio and ffmpeg plugins. Could you sense what else could be done?
~\Anaconda\envs\tensorflow_env\lib\site-packages\imageio_ffmpeg_io.py in read_frames(path, pix_fmt, bpp, input_params, output_params)
148 err2 = log_catcher.get_text(0.2)
149 fmt = "Could not load meta information\n=== stderr ===\n{}"
--> 150 raise IOError(fmt.format(err2))
151 elif "No such file or directory" in log_catcher.header:
152 raise IOError("{} not found! Wrong path?".format(path))

eval_cnn_block_frame_flow.py 35
chkpt = torch.load(model_dir, map_location=lambda storage, loc: storage)

serialization.py 579 load
with _open_file_like(f, 'rb') as opened_file:

serialization.py 230 _open_file_like
return _open_file(name_or_buffer, mode)

serialization.py 211 init
super(_open_file, self).init(open(name, mode))

FileNotFoundError:
2
No such file or directory

Hi,

This is not actually an issue.

I could successfully run datautils.py of the main folder and could generate pickle files corresponding to train, test and validation data for CNN block frame and CNN block frame flow.

I just wanted to know if i want to use keras to train my model how will i load the train,test and val data corresponding to flows if i want train my optical flows(temporal stream)

that is, after loading as below

train_set = BlockFrameFlowDataset(directory, "train")
dev_set = BlockFrameFlowDataset(directory, "dev")
train_set.zero_center(train_set.mean)
dev_set.zero_center(train_set.mean)

train_files, test_files, train_targets, test_targets = train_test_split(......) ??

Any help is highly appreciated.

Thank you

it's not issue but how i can use trained model to detect behavior with opencv ?

when you save the datasets, I think you can not guarantee the frames are always continue

currently, we can test list of videos and get average accuracy. I want to feed the single video file and get a result.

Loading Dataset
Start training
Traceback (most recent call last):
File "train_cnn_single_frame.py", line 59, in
validate=True, resume=resume, use_cuda=cuda)
File "D:\Action Recognition\KTH Program\KTH-Action-Recognition-master\KTH-Action-Recognition-master\main\train_helper.py", line 109, in train
loss = criterion(outputs, labels)
File "C:\Users\Meghna\Anaconda\envs\tensorflow_env\lib\site-packages\torch\nn\modules\module.py", line 489, in call
result = self.forward(*input, **kwargs)
File "C:\Users\Meghna\Anaconda\envs\tensorflow_env\lib\site-packages\torch\nn\modules\loss.py", line 904, in forward
ignore_index=self.ignore_index, reduction=self.reduction)
File "C:\Users\Meghna\Anaconda\envs\tensorflow_env\lib\site-packages\torch\nn\functional.py", line 1970, in cross_entropy
return nll_loss(log_softmax(input, 1), target, weight, None, ignore_index, None, reduction)
File "C:\Users\Meghna\Anaconda\envs\tensorflow_env\lib\site-packages\torch\nn\functional.py", line 1790, in nll_loss
ret = torch._C._nn.nll_loss(input, target, weight, _Reduction.get_enum(reduction), ignore_index)
RuntimeError: Expected object of scalar type Long but got scalar type Byte for argument #2 'target'

the final accuracy obtained from the uploaded model？
My accuracy about block+flow is far from 90%

Hi,

When I run your code, there is no problem with data_utils.py and data_utils.py. But I can not run eval_cnn_block_frame_flow.py sucessfully.

Loading dataset
Loading model
Traceback (most recent call last):
File "eval_cnn_block_frame_flow.py", line 35, in
chkpt = torch.load(model_dir, map_location=lambda storage, loc: storage)
File "/home/yuanpeng/anaconda3/lib/python3.6/site-packages/torch/serialization.py", line 259, in load
f = open(f, 'rb')
FileNotFoundError: [Errno 2] No such file or directory: ''

Hi Khoi,

Simple question here, please direct me on how I can try out my models on a sample video to detect human activity.

Can you tell me how to run the cnn+of with a smaller training set so that my system doesnt crash