Playful interface for the COCO dataset.
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

1.3 KiB
Server for the web interface
Create an svg file per image, with classes on the shapes according to their classes
Create svg frames for a category. Ordered by the area of the shapes.
Find and download the images with only one object in them.
Turn a COCO json file (eg instances_val2017.json) into a database format (eg coco_train.db)

Build array of images sorted by size:

python --annotations ../../datasets/COCO/annotations/instances_train2017.json --output zoom --category_id 18

Turn into png

cd zoom/dog
for file in *.svg; do inkscape -z -f "${file}" -w 640 -e "../dog_png/${file}.png"; done

Turn png into mp4

cd ../dog_png
#ffmpeg -r 1 -i %d_*.png -pix_fmt yuv420p bloch2.mp4
ffmpeg -f image2 -pattern_type glob -i '*.png' ../dog.mp4

To run as server:

cp plottingdata_coco.service /etc/systemd/system/
systemctl daemon-reload
systemctl enable plottingdata_coco.service
systemctl start plottingdata_coco.service
rsync . --exclude zoom --exclude venv --exclude archive -av  --exclude shapes --exclude lonely --exclude .git --exclude __pycache__ --info progress2