| 14 Aug 2019 |
|  high_octane changed the history visibility to "world_readable" from "shared". | 15:57:46 |
| 15 Aug 2019 |
|  doremi joined the room. | 00:02:51 |
| high_octane | Hi, doremi! Welcome to the developer discussion room. I'm currently getting things organized and set up. Specifically, I'm working on the README page for the repo right now. I also wrote a build script (bash) because I'm not a big fan of working with make.
I made a GitHub account, as well as a GitLab account. I really prefer to use GitLab, but GitHub has a larger userbase. I guess that means I'm going to end up maintaining two repos, then. Hmm... Any thoughts? | 03:04:19 |
| doremi | If you prefer GitLab, just use it. Contributors will use it too. I suspect that the vast majority of reference to your project will come from Milovana, not search engine. The user base of either platform is irrelevant, IMO. Also, Micro$oft is the owner of GitHub, and block content in some countries. One more reason to hate them. | 16:42:34 |
| doremi | * If you prefer GitLab, just use it. Contributors will use it too. I suspect that the vast majority of references to your project will come from Milovana, not search engine. The user base of either platform is irrelevant, IMO. Also, Micro$oft is the owner of GitHub, and block content in some countries. One more reason to hate them. | 16:42:55 |
| doremi | * If you prefer GitLab, just use it. Contributors will use it too. I suspect that the vast majority of references to your project will come from Milovana, not search engines. The user base of either platform is irrelevant, IMO. Also, Micro$oft is the owner of GitHub, and block content in some countries. One more reason to hate them. | 16:43:05 |
| high_octane | Good points. It also seems easy enough for GitHub users to log into, and use, GitLab, based on my experience. I use Windows, by the way, but I wouldn't touch Windows 10 with 1,000 ft pole. lol! | 17:18:08 |
| high_octane | I'll be honest, the program doesn't really do much at this point. It can take a video as input and "process" each frame as quickly as it can. No real processing occurs at this moment, however. I need to further understand OpenCV's API before I can proceed. | 17:47:06 |
|  adjones joined the room. | 22:33:30 |
| adjones | Hey, just a short question. Which programming language do u use? I got some experience with machine learning so this project sounds interesting. I dont know yet if I am be able to participate... | 22:35:58 |
| high_octane | This project will be written in C++. I'm not the most familiar with C++, as I'm usually a C programmer. I have some Python and Bash experience as well. Since OpenCV's API is written in C++, that is what I will have to use, but my code contributions will feel very C-like. Do you have prior experience working with OpenCV, adjones? | 23:58:29 |
| 16 Aug 2019 |
| adjones | Yes I have some experience with opencv as well as training of neural networks. I mainly worked with opencv in Python, which I would prefer because of its popularity for machine learning (Tensorflow, Keras...). Anyway, its your choice and C++ is also fine. | 07:11:24 |
| adjones | Short information: I dont have time this weekend, so I wont reply/read messages till next monday. | 07:24:20 |
| high_octane | Okay, whenever you have time is fine. I'm busy cleaning up the codebase and adding the ability to extract media information right now. I want it to be somewhat usable by the time I create the repo and push the first commit, and that might not be for a few more days. | 07:49:38 |
| 21 Aug 2019 |
| high_octane | Hey, everyone! Sorry about the delays. I've been a bit preoccupied lately. I was going to use FFmpeg's avformat to extract some basic video metadata, but learning how to interact with its API is a project in its own right. I've decided to just write my own minimal mp4 parser (and I'll write a parser for the other common containers later). Once I've done that, I'll link you guys to the repository and update the forum thread. | 07:42:14 |
| 22 Aug 2019 |
| doremi | Yeah, these programs are full of libraries, and a knowledge of them all seems to always be a requirement. <facepalm> Could your share some references to simple mp4 frame handling theory/library? Sometimes, I would like co create some video effect to automate what I do in a video editing software, but the video file format isn't the most documented feature, or hidden behind proprietary doors. | 17:07:52 |
| doremi | * Yeah, these programs are full of libraries, and a knowledge of them all seems to always be a requirement. <facepalm> Could you share some references to simple mp4 frame handling theory/library? Sometimes, I would like co create some video effect to automate what I do in a video editing software, but the video file format isn't the most documented feature, or hidden behind proprietary doors. | 17:09:03 |
| doremi | * Yeah, these programs are full of libraries, and a knowledge of them all seems to always be a requirement. <facepalm> Could you share some references to simple mp4 frame handling theory/library? Sometimes, I would like to create some video effects to automate what I do in a video editing software, but the video file format isn't the most documented feature, or hidden behind proprietary doors. | 17:09:56 |
| 23 Aug 2019 |
| high_octane | I'm using the ISO mp4 spec as a resource for my parser. I found it here: http://l.web.umkc.edu/lizhu/teaching/2016sp.video-communication/ref/mp4.pdf | 10:36:04 |
| high_octane | I haven't looked into how to process frames, as I'm only interested in the metadata. What video editing software do you use, doremi? | 10:37:53 |
| 24 Aug 2019 |
| doremi | Thanks! Seems very complex, but seems well documented. I use Vegas. I use zooming and panning in a 4K video for a personal project, and having to key almost every frame always doing the same thing is a necessary thing, but a pure waste of time. It would be so much productive to click a few point defining the area that I need, and the software does the crunching automatically. | 09:33:09 |
| doremi | * Thanks! Seems very complex, but seems well documented. I use Vegas. I use zooming and panning in a 4K video for a personal project, and having to key almost every frame always doing the same thing is a necessary thing, but a pure waste of time. It would be so much productive to click a few points defining the area that I need, and the software does the crunching automatically. | 09:34:45 |
| 25 Aug 2019 |
| doremi | A few OpenSource APIs here: https://www.matroska.org/downloads/linux.html | 21:24:49 |
| high_octane | Thanks for the link, doremi. The Matroska format is beautifully documented on their website, as well. I'll add support for that at a later time, as I'm already holding things up enough as it is. The parser for mp4 is almost done. Yes, I know. I am a slug. Sorry! âšī¸ | 22:31:32 |
| 29 Aug 2019 |
| high_octane | Wow! I just caused a massive bug by accidentally calling fseek() instead of fseeko()! _FILE_OFFSET_BITS=64 is defined, so fseeko() uses a 64-bit offset. | 00:12:20 |
| 31 Aug 2019 |
| high_octane | Okay! The repo is now up and visible on GitLab. I'll update my post on Milo now with a link to it. đ | 03:11:21 |
| high_octane | Alright, everyone! I think the first thing we should try to tackle, ML-wise, is the identification of the beat meter. What are your thoughts? | 05:13:37 |
| high_octane | I might as well post the repository link here as well: https://gitlab.com/high_octane/chext | 17:26:27 |
| 1 Sep 2019 |
| adjones | I think it is necessary to check multiple video frames, as most beatmeters are not visible during breaks or at the beginning.
For the detection I thought about using some color or simple filter hints, but the problem is that there are many different variations (bubbles, waveforms, different positions,...). So maybe train some small CNN?? | 14:23:03 |
