I indexed 669 GB of my GoPro videos using my M1 Max computer and local ML models
I indexed 669 GB of my GoPro videos using my M1 Max computer and local ML models
我使用 M1 Max 电脑和本地机器学习模型索引了 669 GB 的 GoPro 视频
TLDR: I had 2,207 GoPro videos, and I need to rewatch them to find interesting moments from my cycling journey. I built a project to index them locally on my M1 Max using open-source ML models, search for those moments, and send the best clips straight to my DaVinci Resolve timeline. I indexed 628 videos (668.68 GB, 15h 13m 18s of footage duration), more details in the metrics table in the last section of this article.
简而言之：我手头有 2,207 段 GoPro 视频，我需要重温它们以寻找骑行旅程中的精彩瞬间。我构建了一个项目，利用开源机器学习模型在我的 M1 Max 上进行本地索引，搜索这些瞬间，并将最好的片段直接发送到我的 DaVinci Resolve 时间轴中。我索引了 628 段视频（共 668.68 GB，时长 15 小时 13 分 18 秒），更多详细信息请参阅本文最后一节的指标表。
asenna: Funny this is almost EXACTLY what I did a few days ago on the same machine using very similar techniques and was on the front-page of HN as well. I wasn’t familiar with your project though, interesting stuff. I’m trying to add more photography related features to Framedex but yeah there’s so much we can do locally, exciting times.
asenna: 有趣的是，这几乎和我几天前在同一台机器上使用非常相似的技术所做的事情一模一样，而且也上了 HN 的首页。不过我之前不了解你的项目，很有意思。我正在尝试为 Framedex 添加更多与摄影相关的功能，确实，我们可以在本地做的事情太多了，令人兴奋的时代。
iliashad: That’s great, I checked your article when it was in front page because someone mentioned my project in the comments. Good job for the article and the project. That’s great, yes local models are getting better and better.
iliashad: 太棒了，当你的文章上首页时我看了，因为有人在评论区提到了我的项目。你的文章和项目做得很好。确实，本地模型正变得越来越强大。
justinram11: Something I’ve enjoyed more than I expected is Google and Apple photos sending me photo memories and compilations of various things in my life and my kids lives over the last decade. I’m really bullish on taking more video of my kids, with the thought that it will become easier and easier for AI to put them into little compilations I can enjoy later.
justinram11: 有件事让我比预想中更享受，那就是 Google 和 Apple Photos 会给我推送过去十年里关于我生活和我孩子生活的照片回忆和合集。我非常看好拍摄更多关于孩子的视频，因为我认为 AI 将来会越来越容易把它们整理成我可以随时欣赏的小合集。
iliashad: That’s good to hear, open source ML models are getting better and better. I did a small experiment to generate a Spotify year in review like video.
iliashad: 很高兴听到这些，开源机器学习模型确实在不断进步。我做了一个小实验，尝试生成类似 Spotify 年度回顾那样的视频。
esjeon: Aha, it makes total sense. This number sounds much more reasonable than “669 GB”, since the actual total size of processed frames would be like 10-30 GB. I’m just curious tho — is there any paying options that can accelerate this kind of process? Just spin up GPU instances?
esjeon: 啊，这完全说得通。这个数字听起来比“669 GB”合理得多，因为处理后的帧总大小可能只有 10-30 GB 左右。我只是好奇——有没有付费选项可以加速这种处理过程？比如直接启动 GPU 实例？
iliashad: The reason why is “669 GB” is the total raw footage size when I’m doing the video processing, I downscaled each frame to 720p to make the video processing much faster and I don’t need full original quality in order to get accurate results. For now, I found that NVIDIA GPU for example RTX 3060 with 12GB Vram was much faster than my M1 Max.
iliashad: 之所以说是“669 GB”，是因为这是我进行视频处理时的原始素材总大小。我将每一帧缩小到 720p 以加快处理速度，因为要获得准确的结果并不需要原始画质。目前我发现，例如配备 12GB 显存的 NVIDIA RTX 3060 显卡比我的 M1 Max 快得多。
robrain: DaVinci 21 has indexing built-in (AI IntelliSearch). Not to diminish the work you did, but this is now available to many users (probably only Studio users since it has AI in the name).
robrain: DaVinci 21 内置了索引功能（AI IntelliSearch）。并不是要贬低你的工作，但现在很多用户都可以使用这个功能了（可能仅限 Studio 用户，因为名字里带有 AI）。
iliashad: Yes, I didn’t look at it. But does it upload your videos to the cloud or process them locally? And does it allow to provide custom faces data to help labeling faces in your videos?
iliashad: 是的，我没关注过那个。但它是把视频上传到云端还是在本地处理？它是否允许提供自定义人脸数据来帮助标记视频中的人脸？
teovall: The AI features in DaVinci Resolve are all processed locally. It does not currently have face tagging.
teovall: DaVinci Resolve 中的 AI 功能全部是在本地处理的。它目前还没有人脸标记功能。
robrain: Haven’t tried it yet, and I don’t know if it matches OP’s requirements, but the blurb says “You can even search for individual faces”.
robrain: 我还没试过，也不知道它是否符合楼主的要求，但简介里说“你甚至可以搜索特定的人脸”。