Open cuichangrui opened 3 years ago
Hi, are you trying to train or test the model? You might need to download the our pretrained model and put in the correct path if you are just testing. Note that the model I provide gives slightly different results for each action (compared to the reported results) but similar average error.
嗨,您是要训练还是测试模型?如果您只是进行测试,则可能需要下载我们的预训练模型并输入正确的路径。请注意,我提供的模型为每个操作提供的结果略有不同(与报告的结果相比),但平均误差相似。
i was traning this model use "python src/AC_main.py ",but it seems like the h5 file doesn't exist after traning,how can this h5 file be written successfully
thank you for your reply
嗨,您是要训练还是测试模型?如果您只是进行测试,则可能需要下载我们的预训练模型并输入正确的路径。请注意,我提供的模型为每个操作提供的结果略有不同(与报告的结果相比),但平均误差相似。
hi,i had traning the data,and get the h5 file,but when i visualize the predicted results,there has a error"OSError: Unable to open file (unable to open file: name = '../CNNAdTrain_GANWEIGHT0.010000_Sampling0.950000_samples.h5', errno = 2, error message = 'No such file or directory', flags = 0, o_flags = 0)".i has changed the path ,but i noticed in the code ,there has a file named CNNAdTrain_GANWEIGHT0.010000_Sampling0.950000_samples.h5,but i didn't get the h5 file after traning,how can i get this file ? thank you very much
Hi, you might need to run the inference step to generate the samples by setting the 'is_sampling' to True
Hi, you might need to run the inference step to generate the samples by setting the 'is_sampling' to True
hi,i tried modify the "is_sampling" to true,and then use "python src/AC_main.py" run the program,but i got an error "ValueError: Dimensions must be equal, but are 54 and 70 for 'VAESkeleton/add_1' (op: 'Add') with input shapes: [?,1,54,1], [?,1,70,1]. ".Is there any problem with my operation? or can you give me the file named "CNNAdTrain_GANWEIGHT0.010000_Sampling0.950000_samples.h5".thank you for a lot
Which dataset are you training on? Note that the dimension for the human3.6 and CMU dataset is different, namely 54 for human3.6 and 70 for CMU. You might need to specify the dataset if you are using the CMU dataset because the default is human3.6. The other thing is that the model we provide is for human3.6, which can not be applied to CMU directly.
您正在训练哪个数据集?请注意,human3.6和CMU数据集的维度不同,即human3.6和CMU的维度分别为54和70。如果使用的是CMU数据集,则可能需要指定数据集,因为默认值为human3.6。另一件事是我们提供的模型是针对human3.6的,不能直接应用于CMU。
i was traning human3.6m use "python src/AC_main.py", visualize use "python src/forward_kinematics.py"
您正在训练的数据集?请注意,human3.6和CMU数据集的维度不同,即human3.6和CMU的维度分别为54和70。如果使用的是CMU数据集,则可能需要指定数据集,因为原因是人类3.6。另一件事是我们提供的模型是针对人类3.6的,不能直接取代CMU。
我正在使用“ python src / AC_main.py”转换human3.6m,可视化使用“ python src / forward_kinematics.py”
我正在使用“ python src / AC_main.py”转换human3.6m,可视化使用“ python src / forward_kinematics.py”.i only trained human3.6m,and then i use "python src / AC_main.py --is_sampling True" for get the file "CNNAdTrain_GANWEIGHT0.010000_Sampling0.950000_samples.h5",and then i get the error "ValueError: Dimensions must be equal, but are 54 and 70 for 'VAESkeleton/add_1' (op: 'Add') with input shapes: [?,1,54,1], [?,1,70,1]. " Is there any problem with my operation? i need the h5 file for human3.6m.
Oh sorry for the confusion, you might need to change 70 to 'inputDimension' here
Oh sorry for the confusion, you might need to change 70 to 'inputDimension' here
so sorry for asking so much,but 'inputDimension' in the AC_main.py is 70,do not need to change.is there other method for this problem? thank you
Hi, do you try that? The inputDimension should be 54 for human3.6.
Hi, do you try that? The inputDimension should be 54 for human3.6.
hello,i changed the 70 to 'inputDimension',and then use 'python src/AC_main.py --is_sampling True' try to get the file '../CNNAdTrain_GANWEIGHT0.010000_Sampling0.950000_samples.h5',but there has a error 'TypeError: InferenceSample() missing 1 required positional argument: 'iter'. It seems like
'when it run at here ,there lacks the iter
Hi, the 'iter' is to just used to name the folder you will save the sample files, you can give it any number you like. For examples, you can do this: . Note that you need to specify '--checkpoint ' to 24000 if you are using the pretrained model, which means that we restore from the 24000 iteration.
嗨,“ iter”只是用来命名您将保存示例文件的文件夹,您可以给它指定任何数字。例如,您可以执行以下操作: 。请注意,如果您使用的是预训练模型,则需要将'--checkpoint'指定为24000,这意味着我们将从24000迭代中恢复。
thank you so much
OSError: Unable to create file (unable to open file: name = './samples/CNNAdTrain_GANWEIGHT0.010000_Sampling0.950000WindownLength20-2000.h5', errno = 2, error message = 'No such file or directory', flags = 15, o_flags = c2)
Has someone else tried doing so successfully? Any help would be great. Thanks in advance.