search

shuuchen / video_autoencoder

Video lstm auto encoder built with pytorch. https://arxiv.org/pdf/1502.04681.pdf
GNU General Public License v3.0
40 stars 4 forks source link

Movement analysis file #1

Open mortezamg63 opened 4 years ago

mortezamg63 commented 4 years ago

Hello,

Could you please let me know how get "move_vectors_0817_32_all_2lstm.csv"? Also, what is the content of this file?

Thank you

shuuchen commented 4 years ago

I forget where "move_vectors_0817_32_all_2lstm.csv" appears, would you show me the code ?

You can generate the input tensors using Img2Vec as shown in lstm_autoencoder.nbconvert.ipynb

All the images are in pregnant folder

J-sda commented 9 months ago

Hello sir,

I am very appreciate with the code that you have provided. i have some question regarding with the Movement analysis file. What is the "move_vectors_0817_32_all_2lstm.csv" and how did you get it. I noticed that in this file have something related to the movement of the cow

shuuchen commented 9 months ago

Hi,

Thank you for your interest.

The file could be viewed as the latent features of some consistent video frames, with different colors showing the movement of objects inside the video. You can get a similar file by extracting latent features of the autoencoder.

2024年1月16日(火) 19:30 J-sda @.***>:

Hello sir,

I am very appreciate with the code that you have provided. i have some question regarding with the Movement analysis file. What is the "move_vectors_0817_32_all_2lstm.csv" and how did you get it. I noticed that in this file have something related to the movement of the cow

— Reply to this email directly, view it on GitHub https://github.com/shuuchen/video_autoencoder/issues/1#issuecomment-1893468164, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADFCC2P5RPVYAGLDFBMSITDYOZJDHAVCNFSM4OWHD2FKU5DIOJSWCZC7NNSXTN2JONZXKZKDN5WW2ZLOOQ5TCOBZGM2DMOBRGY2A . You are receiving this because you commented.Message ID: @.***>

J-sda commented 9 months ago

Thank you for your response sir.

Base on your response does it mean I need to cluster out the hidden space from the encoder during training and apply the soft-max function. I'm very new to this field, sorry for my silly question. Thank you