Image Captioning Project

#1

Sir, Approx What time will it take if i train my model on my
system (8GB RAM)?

#2

Hey manoj,
it basically depend if you have a GPU or CPU.
On 1050ti 4gb GPU it took around 1 hour to train the model.
On CPU it might take more than 10hours. But you can save model after every epoch , and continue training next day.

Thanks :slight_smile:

#3

I hope I’ve cleared your doubt. I ask you to please rate your experience here
Your feedback is very important. It helps us improve our platform and hence provide you
the learning experience you deserve.

On the off chance, you still have some questions or not find the answers satisfactory, you may reopen
the doubt.

#4

Sir why my encoded_train_feature.pkl file size is only 8.42 kb while Prateek bhaiya’s file has 49.7mb?
Help me please

#5

You might have some issues while writing to the file.
Make sure encoding_train dictionary is correct means (it should have the data)

with open('myfile.pkl', 'wb') as f:
    pickle.dump(mydict, f)
#6

Sir when model predicts how it is stopped from generating any further word. Replay ASAP.

#7

Model can predict any number out of 1848, which corresponds to a single word (mapping can be found using dictionary idx_to_word), and remember 1847 -> startseq and 1848 -> endseq.
Now when model is predicting any number it is mapped with a word first and then this word is appended to the original sentence,
Let’s assume you start and gives input startseq the model predicts a word “dog”, It will be appended to original sentence. Now sentence becomes “startseq dog”, this partial sentence goes as input, and model predict a word a word “dog” and append to sentence, this goes on until we find the word endseq, which literally means this is the end of the sentence.

input output
startseq dog
startseq dog is
startseq dog is running
startseq dog is running endseq

As soon as get the output endseq we will stop predicting next word.

#8

Sir, What does ‘feature extraction’ mean here and what are ‘features’ in Image caption Project?
Reply ASAP.

#9

Features in images means the important pixels or group of pixels that has higher importance than others.
While features in text, means the words each word is a different feature