olivermaerz/landmark2019

Landmark Recognition 2019

Google Landmark Recognition 2019

Tips:

Neither OpenOffice/LibreOffice nor Google Sheets seem to be able to handle the large csv files. So you can either split them: For example to upload them to Google Sheets split them with split -d -l 200000 --additional-suffix=.csv train.csv train.part. and split -d -l 100000 --additional-suffix=.csv train_attribution.csv train_attribution.part. Or use an application that supports large csv files like https://www.csvexplorer.com/ (web based) or http://openrefine.org/ (java app).

How to mount a GCS bucket on an VM instance (running Ubuntu):

First install gcsfuse:

# install gcs
export GCSFUSE_REPO=gcsfuse-`lsb_release -c -s`
echo "deb http://packages.cloud.google.com/apt $GCSFUSE_REPO main" | sudo tee /etc/apt/sources.list.d/gcsfuse.list
curl https://packages.cloud.google.com/apt/doc/apt-key.gpg | sudo apt-key add -
sudo apt update && sudo apt install gcsfuse -y

Then make the directory where you want mount the bucket:

mkdir bucket

Login to Google account:

gcloud auth application-default login

Finally mount the bucket (the GCS bucket's name is landmark-traing)

gcsfuse landmark-training bucket/