Youtube-8m on Nyghtowl

Comparing AI Platform Machine Types using YouTube-8M

Tue, 19 May 2020 18:01:52 +0000

When training a neural net model, time is of the essence. This is why different machine configurations including GPUs, TPUs and multiple servers are utilized.

I’ve been exploring the YouTube-8M project for the last couple months and there are previous posts about the project, the video dataset, the algorithms and how to run them in Cloud. For this post, I trained the two algorithms from the getting started code on different AI Platform standard machine configurations to see how they compared. AI Platform provides a number of scale tiers that are established configurations of different machine types and number of machines to run jobs.

YouTube-8M on AI Platform

Fri, 08 May 2020 00:18:34 +0000

YouTube-8M on AI Platform

Continuing the YouTube-8M exploration and blog series, this post walks through how to use AI Platform to train, evaluate and run predictions for the this dataset. Not surprising, it sets up servers faster than the server I manually configured.

The posts prior to this one provide an overview of the YouTube-8M project, data and computer vision modeling. This research has been used to further computer vision in relation to video datasets over the last several years.

YouTube-8M Training & Inference

Thu, 30 Apr 2020 04:51:54 +0000

Computer Vision | Video Understanding

Continuing on the previous YouTube-8M Dataset post, this one covers model training using what is provided in the getting started section of the GitHub repo. The goal of the models that are covered are to search for a specific moment within a video, which is called temporal concept localization.

In the past, metadata was used to search for videos. These newer models enable classifying specific segments in the video at a specific timestamp where those topics appear. For example the models can help identify in the video all the points where there is chocolate, someone is sleeping or someone is ice skating.

YouTube-8M Dataset

Wed, 11 Mar 2020 20:36:18 +0000

Computer Vision | Video Understanding

YouTube-8M is a project that was developed by Google AI/Research in 2016 to drive innovations and advancement in computer vision, representation learning and video modeling architectures at a large scale.

I’ve been exploring this dataset and example code for a couple weeks and this post summarizes the dataset origin, structure and where to find it. I also share initial exploratory steps that are posted in many places on Kaggle.