GCP Quickstart

To get started using this repo quickly using a Google Cloud Platform (GCP) Deep Learning Virtual Machine (VM) follow the instructions below. New GCP users are eligible for a $300 free credit offer. Other quickstart options for this repo include our Google Colab Notebook and our latest Docker Image.

1. Create VM

Select a PyTorch >=1.3 image from the GCP marketplace, check 'install GPU driver automatically', and select a 100 GB SSD for sufficient I/O speed. All dependencies are included in the preinstalled Anaconda distribution.

2. Setup VM

Clone repo:

git clone https://github.com/ultralytics/yolov3

Download COCO dataset:

bash yolov3/data/get_coco_dataset_gdrive.sh

3. Start

cd yolov3 and:

• Train: python3 train.py
• Test: python3 test.py
• Detect: python3 detect.py

4. Optional Extras

Add 128GB of swap memory (to --cache large datasets).

sudo fallocate -l 128G /swapfile
sudo chmod 600 /swapfile
sudo mkswap /swapfile
sudo swapon /swapfile
free -h  # check memory