Use Pyscenedetect Dynamically In Program

The pydantic schema can be used to describe a class in Python. In this case, the Description class is defined as a base model. The summary provided describes the purpose and functionality of the pyscenedetect library, which assists with dynamically detecting scenes in videos while accounting for possible issues like splicing and video gradients. A code snippet demonstrating how to use this library and FFmpeg to identify and segment a video into scenes is also provided.

对于单纯拼接起来的视频 这个算法就如同手术刀一样精准

首尾有可能有一两帧看起来不太对 但是可以通过调节start和end来修正

警惕频繁转场的视频 它们可能是属于同一个小片段的 但是如果不打乱顺序有可能会触发版权识别问题

即使选出来了可以使用的片段 对于同一个视频的制作过程 依旧要隔一段时间采样 比如两个片段间隔至少5秒 不要单纯的把所有片段一次性提取出来 避免内容重复和版权问题

当然对于有渐变 转场的视频 可能需要用其他的检测方法

1
2
3
4
5
6
7
8
9
10
11
12
13
from scenedetect import open_video, SceneManager, split_video_ffmpeg
from scenedetect.detectors import ContentDetector
from scenedetect.video_splitter import split_video_ffmpeg
def split_video_into_scenes(video_path, threshold=27.0):
# Open our video, create a scene manager, and add a detector.
video = open_video(video_path)
scene_manager = SceneManager()
scene_manager.add_detector(
ContentDetector(threshold=threshold))
scene_manager.detect_scenes(video, show_progress=True)
scene_list = scene_manager.get_scene_list()
split_video_ffmpeg(video_path, scene_list, show_progress=True)

Comments

▲