From 043e06cde21951354b3277eff8c287c9452b0474 Mon Sep 17 00:00:00 2001 From: "Antoine Miech \"WILLOW" Date: Wed, 7 Jun 2017 18:49:48 +0200 Subject: [PATCH] updated readme --- README.md | 20 +++++++++++++++++--- 1 file changed, 17 insertions(+), 3 deletions(-) diff --git a/README.md b/README.md index 12f6efc..bcbc7ca 100644 --- a/README.md +++ b/README.md @@ -1,9 +1,23 @@ # Youtube-8M-WILLOW This is the solution of the kaggle Youtube-8M Large-Scale Video Understanding challenge winners. +For more details about our models, please read our arXiv paper: ... . -The code is build on top of the Google Youtube-8M starter code (https://github.com/google/youtube-8m) +The code is built on top of the Google Youtube-8M starter code (https://github.com/google/youtube-8m) Please look at their README to see the needed dependencies to run the code. -On top of this, you will only need to have the pandas python library installed. +You will additionally only need to have the pandas python library installed. -The models have been run on NVIDIA TITAN X 12 GB GPU. +Hardware requirement: Each model can be run on a single NVIDIA TITAN X 12 GB GPU. Some of the trained +do not fit with a GPU with at least 10GB of memory. In that case, please do not modify the training batch size +of these models as it might effect the final results. + +Our submitted model is a weighted ensemble of 25 models. +However for the sake of simplicity, this repot will contain a much more simple ensemble of +7 models that are enough to reach the first place with a significant margin. + + + + + + +Antoine Miech