search

Shreyz-max / Video-Captioning

Video Captioning is an encoder decoder mode based on sequence to sequence learning
https://shreyz-max.github.io/Video-Captioning/
119 stars 35 forks source link

Features extraction #10

Open LuninaPolina opened 1 year ago

LuninaPolina commented 1 year ago

Hi! Thanks a lot for you project!

I have a problem with features extraction (extract_feature.py file).

Running prediction on npy file presented in testing_data folder produces different result from applying features extraction and then predicting text caption. For example, on video 3qqEKTPxLNs_1_15 first scenario produces output "a baby is playing with the ball" (which is correct) while the second one outputs "a woman is making a woman" (which is just strange).

I looked into this issue, because it seemed relevant to my problem, however, I don't have any files starting with ".', so, removing them didn't help.

Could you please help me with my problem, because I need to generate text description on custom videos that aren't presented in msvd dataset.

the-ntg commented 1 year ago

Yaa I am facing same issue

Hi! Thanks a lot for you project!

I have a problem with features extraction (extract_feature.py file).

Running prediction on npy file presented in testing_data folder produces different result from applying features extraction and then predicting text caption. For example, on video 3qqEKTPxLNs_1_15 first scenario produces output "a baby is playing with the ball" (which is correct) while the second one outputs "a woman is making a woman" (which is just strange).

I looked into this issue, because it seemed relevant to my problem, however, I don't have any files starting with ".', so, removing them didn't help.

Could you please help me with my problem, because I need to generate text description on custom videos that aren't presented in msvd dataset.

nkumawat34 commented 1 year ago

I am also facing same issue have you find any solution.