guangmingzhu / attentionconvlstm Goto Github PK

Thanks for your code. Do you know whether the keras.layers.GatedConvLSTM2D is changed into keras.layers.GatedConvLSTM2D? And i also can't find keras.layers.GatedConvLSTM2D function in keras.Can you provide some reference?Thank you.
Another question: When i add lstm in i3d code but the loss is nan,do you know why?

Hi there

Great work on this repo / paper! I would like to extend it possibly.

Would it be possible for you to upload the weights to another location?

I am not able to signup for Baidu drive as it requires me to have a Chinese mobile number. Plus I do not speak Chinese so using the app will be difficult for me.

Could you upload the weights somewhere else, like Google Drive (or anywhere really)?

Thanks in advance
Liam Schoneveld

一天了才训了3个epoch正常吗？GPU显存一直占满，但利用率时常降为会为0
能怎么提提速吗？
谢谢

Dear,

when I run the train example , the problem occured like this:

2019-04-25 16:08:15.995839: E tensorflow/stream_executor/cuda/cuda_dnn.cc:359] could not create cudnn handle: CUDNN_STATUS_INTERNAL_ERROR
2019-04-25 16:08:15.995915: E tensorflow/stream_executor/cuda/cuda_dnn.cc:326] could not destroy cudnn handle: CUDNN_STATUS_BAD_PARAM
2019-04-25 16:08:15.995941: F tensorflow/core/kernels/conv_ops_3d.cc:395] Check failed: stream->parent()->GetConvolveAlgorithms(&algorithms)
Aborted

thank you!

I use tensorfloe-gpu-1.2.0 and cudnn-5.0

周老师，您好！我在我自己的笔记本上训练模型，发现速度特别慢。所以我想请问一下，您们是在什么配置电脑上训练的，花了大约多久？

我现在拿到了Iso_GD数据集,但是没有对应的视频文件转换为图片的脚本,没有办法有效的开展网络的训练和测试.

The program did not finish running an epoch, a path error occurred.I checked the dataset and there was no problem.I made the data set smaller, and it will give an error when executing 49/50, and this problem will also occur when executing 79/80.Thanks！
@GuangmingZhu

hi author, when I run training_.py, I got error -- cannot import name 'keras_modules_injection'. I have already replace init and convolution_recurrent files to my tensorflow 1.20. how to solve it? thanks!

File "/Library/Frameworks/Python.framework/Versions/3.6/lib/python3.6/site-packages/tensorflow/python/keras/applications/inception_v3.py", line 24,in
from tensorflow.python.keras.applications import keras_modules_injection
ImportError: cannot import name 'keras_modules_injection'

your paper only show the result of each modality, how about the fusion result of RGB depth and flow?

你好，请问这个是动态连续帧手势识别项目吗？谢谢！

After read your paper,I wonder how to visualization? Thanks

您好 我可以问一下数据集划分里面的flow-data是什么文件吗 我拿到了IsoGD数据集里面并没有f开头的视频

Hi Dr. Zhu,

Thanks for your contributions. When I used your code to train the model, I found the training accuracy increased a little with the rise of epochs while the validation accuracy seemed not to change at different epochs. But both the training accuracy and the validation accuracy were no more than 0.2, which were too low.

May I ask whether there are any methods to solve this problem?

我看到不同的数据都有不同average_values，请问这个值是怎么确定的呢？尤其是flow数据的average_values值？

inputs模块中没有load_iso_video_list函数
另外，如果想使用jester集训练，对应的函数是什么呢？
谢谢

有什么好的方法能够快速将大量视频转换为图片吗？

I got the error:
ValueError: Only provide the shape OR batch_shape argument to Input, not both at the same time.
I don't know why? Can you help me?Thanks!
@GuangmingZhu

The issue occured when I use Jester date set,I'm sure that the lable list is right.

Traceback (most recent call last):
File "training_res3d_clstm_mobilenet.py", line 126, in
initial_epoch=init_epoch,
File "/home/usr2/anaconda2/lib/python2.7/site-packages/tensorflow/contrib/keras/python/keras/engine/training.py", line 1857, in fit_generator
'or (x, y). Found: ' + str(generator_output))
ValueError: output of generator should be a tuple (x, y, sample_weight) or (x, y). Found: None

I am in torubled for some days,can any body help me ?Thanks so much!!!

Hello,
Congrats, for the wonderful work. I am facing some issue in replicating the results of your paper using the source code provided here.

This portion of the training code

featmap_2d = keras.layers.Reshape((28,28,256), name='clstm_respahe')(clstm2d_2)
is giving the below error

tensorflow.python.framework.errors_impl.InvalidArgumentError: Input to reshape is a tensor with 25690112 values, but the requested shape has 1605632
Explanation:
I have found out the problem, but not able to resolve it.

Input to reshape tensor = (batch_size * L/2 * 28 * 28 * 256) = 25690112 (L=32)
Requested shape = (batch_size * 28 * 28 * 256) = 1605632

keras.layers.Reshape needs same tensor size, but here there is mismatch in the size. Please help me in solving the issue.

Thank You

'tensorflow.contrib.keras.api.keras.layers' has no attribute 'GatedConvLSTM2D'

BSD?

hello, this is the error I encountered, hope to get any help,thank you~
Traceback (most recent call last):
File "training_res3d_clstm_mobilenet.py", line 96, in
feature = res3d_clstm_mobilenet(inputs, seq_len, weight_decay)
File "./networks/res3d_clstm_mobilenet.py", line 199, in res3d_clstm_mobilenet
features = keras.layers.Reshape((seq_len/2,4,4,1024), name='feature_reshape')(features)
File "/home/lab226/anaconda3/envs/tf/lib/python3.6/site-packages/tensorflow/contrib/keras/python/keras/engine/topology.py", line 396, in call
output = super(Layer, self).call(inputs, **kwargs)
File "/home/lab226/anaconda3/envs/tf/lib/python3.6/site-packages/tensorflow/python/layers/base.py", line 450, in call
outputs = self.call(inputs, *args, **kwargs)
File "/home/lab226/anaconda3/envs/tf/lib/python3.6/site-packages/tensorflow/contrib/keras/python/keras/layers/core.py", line 402, in call
return K.reshape(inputs, (-1,) + tuple(target_shape))
File "/home/lab226/anaconda3/envs/tf/lib/python3.6/site-packages/tensorflow/contrib/keras/python/keras/backend.py", line 1892, in reshape
return array_ops.reshape(x, shape)
File "/home/lab226/anaconda3/envs/tf/lib/python3.6/site-packages/tensorflow/python/ops/gen_array_ops.py", line 2619, in reshape
name=name)
File "/home/lab226/anaconda3/envs/tf/lib/python3.6/site-packages/tensorflow/python/framework/op_def_library.py", line 589, in apply_op
param_name=input_name)
File "/home/lab226/anaconda3/envs/tf/lib/python3.6/site-packages/tensorflow/python/framework/op_def_library.py", line 60, in _SatisfiesTypeConstraint
", ".join(dtypes.as_dtype(x).name for x in allowed_list)))
TypeError: Value passed to parameter 'shape' has DataType float32 not in list of allowed values: int32, int64

I found the steps listed very difficult to follow and understand.
especially it is using python 2.7 which is deprecated on Google Colab
could any one tell me how to start with this model?