Add a tool under visual subfolder for command-line shot detection

tyiannak / multimodal_movie_analysis

A Python Library for Multimodal Analysis of Movies and Content-based Movie Recommendation

25 stars 8 forks source link

Add a tool under visual subfolder for command-line shot detection #9

Closed tyiannak closed 3 years ago

tyiannak commented 3 years ago

Description

create a command-line tool that takes a video (or a folder of videos) and performs shot boundary detection
also saves all shots under separate video files named after the __
algorithm: open video, read per frame (as in analyze_visual), then convert to grayscale and then uses shot_change() from analyze_visual. Finally, based on the shot limits it saves shot videos using ffmpeg. e.g. ffmpeg -i large_movie.avi -ss 10 -to 20 large_movie_10_20.avi)

tyiannak commented 3 years ago

Also do some qualitative evalauation (does shot change work on movies)?

apoman38 commented 3 years ago

A new file named shot_generator.py was created based on analy_visual.py

Using the existing function shot_change it was easy to find the shots. Initially, a new movie_time variable was added where the timestamp is stored when the shot changes. The milliseconds were then converted in the form of for this purpose a new function called milisec_to_time was created. At the end, the code additionally prints the number of shots contained in the video, the time at which the shot starts and the moment it ends (in NumPy array format). The end result is the creation of many different videos where their number is equal to the number of shots contained in the video requested by the user.

Please check the code and let me know for bugs or any problems.

P.S Need to install a new library named moviepy

apoman38 commented 3 years ago

I forgot to mention that I added 2 new videos to the dataset folder in case you want to test the code.

tyiannak commented 3 years ago

The code should be much shorter. There's no need to use all code from analyze_visual.py, just do the following:

import analyze_visual
do a minor change in process_video() to also return shot limits
call process_video()
write shot videos based on the returned shot limits The code should be less than 30 lines of code and without duplicates from analyze_visual.py

apoman38 commented 3 years ago

Code refactoring was performed on the shot_generator.py file

The analyze_visual.py file was imported to access all functions and libraries
2 new functions were created (crop_shots, crop_dirs) the first is used in case the user gives as an argument a single video and the second in when a folder is given by the user.

P.S The code is executed in the same way as in the analyze_visual.py file

apoman38 commented 3 years ago

shot_generator.py works properly.