Indexerror
Opened this issue · 22 comments
Dear Dongfeng,
I get an error as : IndexError: index 5 is out of bounds for axis 1 with size 5 when I run the program. I would be most obliged if you can help me to solve this error. Thank you
I had the same error fixed it by changing the number of classes in configuration
I also met this situation,how do you deal with it? Where is the configuration changed?
It is in run_dense_net_3d.py file line no 13 n_classes change it according to your classes
here is the Whole Configuration or training prams, Change them according to your requirements
train_params = {
'num_classes': 5,
'batch_size': 10,
'n_epochs': 70,
'crop_size': (150,100),
'sequence_length': 16,
'initial_learning_rate': 0.1,
'reduce_lr_epoch_1': 30, # epochs * 0.5
'reduce_lr_epoch_2': 55, # epochs * 0.75
'validation_set': True,
'validation_split': None, # None or float
'queue_size': 300,
'normalization': 'std', # None, divide_256, divide_255, std
}
thank you, I have solved the problem
What's the accuracy of your experiment using this code ? Why do I have less than 1% mean accuracy
my dataset is UCF101
@WanliOuyang Just Check the Result While testing Maybe its the normalized Accuracy also you can try other implementation of this Also how many Classes do you have if you have 101 classes then it will take long time to converge
it needs 5 or 6 days to run the code when I use UCF101 dataset with 101 classes, now,the training accuracy and validation accuracy is up to 0.013, so Where was my mistake?
Well it Depends upon the System Specifications Also what are your hardware Specs ?
name: Quadro M4000
major: 5 minor: 2 memoryClockRate:0.7725
total memory: 7.93GiB
free memory: 6.91GiB
Its Fine i mean the Specs are Ok, Actually i ran the model but my data-set were very small and had only 3 classes for that it converged successfully. I don't know Its result on UCF 101 because i haven't run it on UCF101 although the author has given the result on UCF and some other datasets, what you can do is just test the model on the already saved checkpoints, by this way you will get the idea if its working or not. But it Should work you have to wait Run it at-least for 20,000 Epochs
Hello, do you run other datasets on this code? For example, MERL_shopping、KTH.
When training, it appears "data_providers/base_provider.py : 33 : RuntimeWarning : invalid value encountered in divide image=(image-np.mean(image))/np.std(image)
What should I do to solve this problem?
You can set a condition: the denominator is not zero when the code is executed
Thank you. Now, I am training it on UCF101. I set the epochs as 20000, but its accuracy is 0.012 when 11 epochs costing 12 hours.
Is there something wrong?
It must be that the configuration of the parameters is not reasonable, and you can adjust the parameters properly.
Can I get a list of configuration of yours that running on UCF101? I will be appreciate very much to you .
Hello together, i seem to be unable to get UCF101 to converge even with setting like the ones used on the densenet-general branch. KTH seems to be converging instantly though.
I would be intereseted if anyone found a parameter setting for UCF101
How Many Classes I set Because i'm beginner and i have to run this code ( dataset UCF101)
How Many Classes I set Because i'm beginner and i have to run this code ( dataset UCF101)
It should be 101 I think .
How Many Classes I set Because i'm beginner and i have to run this code ( dataset UCF101)
It should be 101 I think .
thanks it solved the problem.
Please can you share me the accuracy that you got for kth (can we talk at skype because i'm beginner and i want more clarification ?)
How Many Classes I set Because i'm beginner and i have to run this code ( dataset UCF101)
It should be 101 I think .
thanks it solved the problem.
Please can you share me the accuracy that you got for kth (can we talk at skype because i'm beginner and i want more clarification ?)
OK. You can send your WeChat id to my E-mail. 1920186582@qq.com