-
Notifications
You must be signed in to change notification settings - Fork 1
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
- Loading branch information
1 parent
c8fb2fa
commit cc284be
Showing
10 changed files
with
609 additions
and
40 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,169 @@ | ||
| +++ | ||
| title = '基于Python与K-Means的自动化视频分类方法' | ||
| date = 2024-09-21T01:43:21.619108+08:00 | ||
| draft = false | ||
| +++ | ||
|
|
||
| > 版权归作者所有，如有转发，请注明文章出处：<https://cyrus-studio.github.io/blog/> | ||
| # __实现过程__ | ||
|
|
||
|
|
||
| 1\. 特征提取：使用预训练的 InceptionV3 模型，从视频的若干帧中提取高维的视觉特征。将每个视频的所有帧特征取平均值，生成一个固定长度的特征向量来表示该视频。 | ||
|
|
||
| 2\. 聚类：通过 K-Means 的聚类结果，每个视频被分配了一个簇标签，代表该视频与哪些视频在特征上最相似。 | ||
|
|
||
| 3\. 分类整理：最后根据簇标签，将视频移动到相应的分类文件夹中，每个文件夹对应一个簇。 | ||
|
|
||
| ## __InceptionV3 模型__ | ||
|
|
||
|
|
||
| InceptionV3 是一种用于图像分类和特征提取的深度学习模型，它是Inception 系列模型的第三个版本，由 Google 在 2015 年提出。 | ||
|
|
||
| 它最初是作为图像分类任务的一个模型，能够将图像分类到 1000 个类别中（如狗、猫、汽车等）。通过去除模型的最后几层（分类部分），可以将 InceptionV3 用作特征提取器。 | ||
|
|
||
| ## __簇__ | ||
|
|
||
|
|
||
| 簇是聚类算法的核心概念，表示数据中相似的子集，目的是将无标签的数据点分组。 | ||
|
|
||
| ## __K-Means__ | ||
|
|
||
|
|
||
| K-Means 是一种常用的无监督聚类算法，它的目标是将数据点分成 K 个簇（Cluster），使得每个簇内的数据点尽可能接近同一个中心（即簇的质心）。 | ||
|
|
||
| 算法的核心思想是通过迭代的方式找到 K 个最优的簇质心，并根据这些质心将数据进行分组。 | ||
|
|
||
| # __源码__ | ||
|
|
||
|
|
||
| ## __1\. 安装依赖库__ | ||
|
|
||
|
|
||
| ``` | ||
| pip install moviepy scikit-learn tensorflow opencv-python | ||
| ``` | ||
|
|
||
| ## __2\. 实现代码__ | ||
|
|
||
|
|
||
| ``` | ||
| import os | ||
| import numpy as np | ||
| import cv2 | ||
| from moviepy.editor import VideoFileClip | ||
| from sklearn.cluster import KMeans | ||
| from tensorflow.keras.applications import InceptionV3 | ||
| from tensorflow.keras.applications.inception_v3 import preprocess_input | ||
| from tensorflow.keras.preprocessing import image | ||
| from tensorflow.keras.models import Model | ||
| from shutil import move | ||
| # 提取视频的帧作为特征 | ||
| def extract_video_features(video_path, model, frame_interval=30): | ||
| video = VideoFileClip(video_path) | ||
| frame_count = 0 | ||
| features = [] | ||
| for frame in video.iter_frames(fps=1): # 以每秒一帧的速度获取帧 | ||
| if frame_count % frame_interval == 0: | ||
| # Resize frame to match model input size (299x299 for InceptionV3) | ||
| img = cv2.resize(frame, (299, 299)) | ||
| img = image.img_to_array(img) | ||
| img = np.expand_dims(img, axis=0) | ||
| img = preprocess_input(img) | ||
| # 提取特征 | ||
| feature = model.predict(img) | ||
| features.append(feature.flatten()) | ||
| frame_count += 1 | ||
| # 取视频的所有帧特征的均值作为视频的最终特征 | ||
| return np.mean(features, axis=0) | ||
| # 批量提取目录下所有视频的特征 | ||
| def extract_features_for_all_videos(input_dir, model, frame_interval=30): | ||
| video_features = [] | ||
| video_files = [] | ||
| for filename in os.listdir(input_dir): | ||
| if filename.endswith(".mp4"): # 你可以根据需要修改文件格式 | ||
| video_path = os.path.join(input_dir, filename) | ||
| print(f"正在处理视频: {filename}") | ||
| features = extract_video_features(video_path, model, frame_interval) | ||
| video_features.append(features) | ||
| video_files.append(filename) | ||
| return np.array(video_features), video_files | ||
| # 对视频进行聚类 | ||
| def cluster_videos(video_features, num_clusters=3): | ||
| kmeans = KMeans(n_clusters=num_clusters, random_state=42) | ||
| kmeans.fit(video_features) | ||
| return kmeans.labels_ | ||
| # 将视频分类到不同的文件夹 | ||
| def classify_videos(input_dir, output_dir, video_files, labels): | ||
| for label, filename in zip(labels, video_files): | ||
| output_folder = os.path.join(output_dir, f"cluster_{label}") | ||
| if not os.path.exists(output_folder): | ||
| os.makedirs(output_folder) | ||
| input_path = os.path.join(input_dir, filename) | ||
| output_path = os.path.join(output_folder, filename) | ||
| move(input_path, output_path) | ||
| print(f"已将视频 {filename} 移动到 {output_folder}") | ||
| # 主函数 | ||
| def main(input_dir, output_dir, num_clusters=3, frame_interval=30): | ||
| # 加载预训练的InceptionV3模型，并去掉顶层的分类部分，只用来提取特征 | ||
| base_model = InceptionV3(weights='imagenet') | ||
| model = Model(inputs=base_model.input, outputs=base_model.get_layer('avg_pool').output) | ||
| # 提取所有视频的特征 | ||
| video_features, video_files = extract_features_for_all_videos(input_dir, model, frame_interval) | ||
| # 对视频进行聚类 | ||
| labels = cluster_videos(video_features, num_clusters) | ||
| # 将视频移动到相应的分类文件夹 | ||
| classify_videos(input_dir, output_dir, video_files, labels) | ||
| # 示例调用 | ||
| input_directory = "path/to/input_videos" | ||
| output_directory = "path/to/output_videos" | ||
| main(input_directory, output_directory, num_clusters=30, frame_interval=30) | ||
| ``` | ||
|
|
||
| ## __3\. 代码说明__ | ||
|
|
||
|
|
||
| 1\. extract_video_features：从每个视频中提取帧，使用 InceptionV3 模型提取每个帧的特征，并最终取所有帧特征的平均值作为该视频的代表特征。 | ||
|
|
||
| 2\. extract_features_for_all_videos：批量提取目录中所有视频的特征。 | ||
|
|
||
| 3\. cluster_videos：使用 K-Means 聚类算法对视频进行分类，将相似的视频聚到一起。 | ||
|
|
||
| 4\. classify_videos：将视频根据聚类结果移动到不同的分类文件夹。 | ||
|
|
||
| 5\. main：主函数，负责加载模型、提取特征、聚类以及将视频分类。 | ||
|
|
||
| ## __4\. 调用说明__ | ||
|
|
||
|
|
||
| 1\. input_directory: 视频所在的输入文件夹。 | ||
|
|
||
| 2\. output_directory: 输出文件夹，程序会根据聚类结果创建不同的文件夹，将相似的视频分类进去。 | ||
|
|
||
| 3\. num_clusters: 要分类的类别数，即希望将视频分为多少类。 | ||
|
|
||
| 4\. frame_interval: 每隔多少帧提取一次特征帧。值越大，提取帧的间隔越大。 | ||
|
|
||
|
|
||
| 源码地址：[https://github.com/CYRUS-STUDIO/classify-videos-kmeans-python](https://github.com/CYRUS-STUDIO/classify-videos-kmeans-python) | ||
|
|
||
|
|
||
|
|
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,96 @@ | ||
| <!doctype html> | ||
| <html lang="zh-cn"> | ||
| <head> | ||
| <title>CYRUS STUDIO</title> | ||
| <link rel="shortcut icon" href="/favicon.ico" /> | ||
| <meta charset="utf-8" /> | ||
| <meta name="generator" content="Hugo 0.131.0"> | ||
| <meta name="viewport" content="width=device-width, initial-scale=1" /> | ||
| <meta name="author" content="John Doe" /> | ||
| <meta name="description" content="Android & Python Developer" /> | ||
| <link rel="stylesheet" href="/blog/css/main.min.08e876a0f4aeb92fb7ca4e4c12d7d6ea16684353b267ad3f4385e180cc91a06b.css" /> | ||
| <link rel="alternate" type="application/rss+xml" href="https://cyrus-studio.github.io/blog/index.xml" title="CYRUS STUDIO"> | ||
|
|
||
|
|
||
|
|
||
|
|
||
|
|
||
|
|
||
|
|
||
| <meta name="twitter:card" content="summary"> | ||
| <meta name="twitter:title" content="CYRUS STUDIO"> | ||
| <meta name="twitter:description" content="Android & Python Developer"> | ||
|
|
||
| <meta property="og:url" content="https://cyrus-studio.github.io/blog/"> | ||
| <meta property="og:site_name" content="CYRUS STUDIO"> | ||
| <meta property="og:title" content="CYRUS STUDIO"> | ||
| <meta property="og:description" content="Android & Python Developer"> | ||
| <meta property="og:locale" content="zh_cn"> | ||
| <meta property="og:type" content="website"> | ||
|
|
||
|
|
||
| </head> | ||
| <body> | ||
| <header class="app-header"> | ||
| <a href="https://cyrus-studio.github.io/blog/"><img class="app-header-avatar" src="/blog/avatar.jpg" alt="John Doe" /></a> | ||
| <span class="app-header-title">CYRUS STUDIO</span> | ||
| <nav class="app-header-menu"> | ||
| <a class="app-header-menu-item" href="/blog/">Home</a> | ||
| - | ||
|
|
||
| <a class="app-header-menu-item" href="/blog/about/">About</a> | ||
| - | ||
|
|
||
| <a class="app-header-menu-item" href="https://github.com/CYRUS-STUDIO">GitHub</a> | ||
| </nav> | ||
| <p>Android & Python Developer</p> | ||
| </header> | ||
| <main class="app-container"> | ||
|
|
||
| <article> | ||
| <h1>CYRUS STUDIO</h1> | ||
| <ul class="posts-list"> | ||
|
|
||
| </ul> | ||
|
|
||
|
|
||
| <ul class="pagination"> | ||
|
|
||
| <li class="page-item"> | ||
| <a class="page-link" href="/blog/page/2/"> | ||
| <svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="icon icon-arrow-left"> | ||
| <title>arrow-left</title> | ||
| <line x1="19" y1="12" x2="5" y2="12"></line><polyline points="12 19 5 12 12 5"></polyline> | ||
| </svg> | ||
| </a> | ||
| </li> | ||
|
|
||
|
|
||
| <li class="page-item"> | ||
| <a class="page-link" href="/blog/"> | ||
| 1 | ||
| </a> | ||
| </li> | ||
|
|
||
| <li class="page-item"> | ||
| <a class="page-link" href="/blog/page/2/"> | ||
| 2 | ||
| </a> | ||
| </li> | ||
|
|
||
| <li class="page-item active"> | ||
| <a class="page-link" href="/blog/page/3/"> | ||
| 3 | ||
| </a> | ||
| </li> | ||
|
|
||
|
|
||
| </ul> | ||
|
|
||
|
|
||
|
|
||
| </article> | ||
|
|
||
| </main> | ||
| </body> | ||
| </html> |
Oops, something went wrong.