Using FFMPEG to Encode/Decode Video for Offline Video-Based Machine Learning

preview_player
Показать описание
FFMPEG is a tool that can be used to both decode and encode video between formats such as .MP4/.MOV to/from individual JPEG images. This technique allows machine learning programs to process the individual frames of a video and even generate entirely new videos.

Here are the three notebooks from this video:

Рекомендации по теме
Комментарии
Автор

This is awesome stuff. I always found ffmpeg difficult to use so thank you for this very clear walkthrough. You’re a great teacher and I always look forward to your videos.

joliver
Автор

You actually did a video on this. Excited to watch

MawaMaverick
Автор

Hey! Waiting for the live video one, great video

davidgeorge
Автор

amazing work always, every video and lecture man.
pleasee more of this topic

luisilabaca
Автор

And jeff gave this to us for free. Amazing

MawaMaverick
Автор

enjoyed ... looking forward to more ... Be safe!

MIGuy
Автор

Hi. Jeff, thanks a lot for posting the AI tutorial. It helps a lot for a beginner like me to learn AI. I have a question, how do we crop images inside of the bounding box of YOLOV3 when the input is a video? So in the end, we have images of each class detected in the video. Would you please make a tutorial about it? Thanks

jokosusilo
Автор

Hi, thank you for the video. I would like to ask you if there is some way to access the DCT coefficients from a compressed video?

efstathiasoufleri
Автор

Thank you for wonderful presentation, Is there a way to start from a user inserted image and begin the morphing from there?

quisaldaoud
Автор

I have a setup where I'm using ffmpeg for encoding and transcoding for live streams but presently experiencing lots of latency despite deployment of CloudFront et all.

Is there a way i can set up ffmpeg to give better performance which in turn will help reduce latency?

OkolieRaymond
Автор

Hi Jeff, what about using ML to strip out a lead singer's voice from an MP3 file to make a karaoke track? As sound is time-series data, I thought it would be interesting but I have no idea if that is possible or if it's too much of a heavy lift for Youtube. I thought I read somewhere that that was possible.

kenchang
Автор

i have a video of a liquid being heated, is there a way to make a neural network that if given a couple of seconds it will predict the next few seconds? will you be working on something similar in this series?

FMH
Автор

can we use ffmpeg for online realtime video processing also?

frazierkok
Автор

Hello Jeff and community,
I am currently using ffmpeg to decode/extract frames from video. It... just works. But its heck slow.
Im having a wrapper where I can use the function GerFrame() which basically pulls the next frame. It is a little faster than watching it in media player but at this rate, it will take 50 Minutes to process 1h of video into images. Is that normal? If so I have to look at other methods than ffmpeg. Apparently, I can watch a video at 4x speed within VLC mediaplayer. So it should not be an issue by hardware speed?
LG,
Julian

bechti
Автор

Can You please say how can I convert mp4 video file frame by frame to json format using ffmpeg or any source code is available?

ayandhar
Автор

It's amazing. Can you please build the model to translate speech from one language to another in video.

krantikumar
Автор

Thanks for this video. I'm interested in CV but NLP is more practical for now, don't you think?!

kenchang
Автор

Hi Jeff,


How would we run this locally rather than on Google Colab? Specifically the command:


from google.colab import files
uploaded = files.upload()


Thank you Jeff!

RedShipsofSpainAgain
Автор

You talk about real time video processing with open CV and augmented reality. I would really like to see a video about AI following an objet on a (phone) screen and providing its position as fast as possible. It's for video games. Thanks.

Lalancelot
Автор

"I like no work on my part." LOL! I've always told my students that we programmers are lazy - not in that we won't do what needs doing, but that we try to do it with a goal of maximizing functionality while minimizing effort. After ll, those cat videos won't watch themselves! :-)

BradHelm
join shbcf.ru