Cars, People, Tents — Vancouver Watching w/ YOLOv5 🚀

All the pieces are in place to train a YOLOv5 🚀. First, let’s validate the training pipeline.

abcli select
yolov5 train coco128 \
    classes=person+bicycle+car+motorcycle+bus+train+truck+bird

This command ingests coco128, crops the classes to the given list, and then runs,

python3 \
    train.py \
    --img 640 \
    --batch 16 \
    --epochs 3 \
    --data $abcli_object_root/$dataset_name/dataset.yaml \
    --weights yolov5n.pt \
    --project $abcli_object_path \
    --workers 0 \
    --name model \
    --device cpu

After ~15 minutes of training, a validation model is trained.

Read more about validation models here.

Next step: train a production model.

Here is a look at the dev env that I use.

left: git, right, top: ec2, bottom: local, bottom: a browser