search

weirme / FCSN

A PyTorch reimplementation of FCSN in paper "Video Summarization Using Fully Convolutional Sequence Networks"
115 stars 33 forks source link

datasets #2

Open YapingZ opened 5 years ago

YapingZ commented 5 years ago

Why the dataset cannot be downloaded？

SinDongHwan commented 5 years ago

Can't you open dataset?

if you can't open, access here for this model. this link is same data.

YapingZ commented 5 years ago

When training the tvsum dataset, the range of index in the data_loader.py file always exceeds the number of datasets, and there is a problem with the code.

SinDongHwan commented 5 years ago

Yes, Maybe you should edit a code in data_loader.py

video=self.data_file['video_'+str(index.item())] to video=self.data_file['video_'+str(index)]

beacuse "index" is just variable stored int data-type.

YapingZ commented 5 years ago

our test code does not get the accuracy rate and the recall rate.

YapingZ commented 5 years ago

Will you improve the test code? Such as visualization, generating video, and getting accurate f-scores

SinDongHwan commented 5 years ago

I would be able to do.. Maybe ^^. I will try to visualize loss and acc graph, generate summarization and calculate f-scores. But i want you to know, i'm not good at programming skills. Thank you^^

YapingZ commented 5 years ago

Do you have WeChat or email? Some questions want to ask you。

SinDongHwan commented 5 years ago

my email is God.DongHwan@gmail.com

if i know answer for you questions. i can.

if is not, i will answer. but i will try to solve.