describe image in English with AI
images/test1.jpg
images/test2.jpg
images/test3.jpeg
./aiid-describe -g imagesimages/test1.jpg → a car driving down a dirt road in the fall
images/test2.jpg → a person standing on top of a sand dune
images/test3.jpeg → a boat is in the water near a city
./setup-venv.sh # nvidia gpu
# or
./setup-venv-rocm.sh # amd gpu./install.sh
# Creates symlinks to shell
# scripts that run the
# python scripts in a venv.usage: aiid-describe.py [-h] [-s] [-c number] [-w] [-wc] [-g] [-b number]
[-lb number] [-o file.csv] [-od file.csv] [-d] [-d2]
input
Describe images in English using BLIP
positional arguments:
input image file or directory with images
options:
-h, --help show this help message and exit
-s, --show show image(s) with caption in title after processing
all
-c, --count number how many captions to generate per image (default: 1)
-w, --words turn captions into keywords that will be sorted
-wc, --words_clean same as -w + remove function words like ('the', 'a',
'an', ...)
-g, --gpu use gpu
-b, --batch number batch size for GPU processing
-lb, --load_batch number
batch size for loading images into memory
-o, --output file.csv
output results to CSV file
-od, --output_delimiter file.csv
delimiter to use when outputing csv file (default:
'|')
-d, --defaults -c 5 -wc -g -b 128 -lb 256 -o output.csv
-d2, --defaults2 -c 5 -wc -g -b 256 -lb 512 -o output.csv./aiid-describe -g images -d(needs manual tweaking to actually work well)
usage: aiid-group.py [-h] [-s] [-g] directory
group images by eng names
positional arguments:
directory dir with images
options:
-h, --help show this help message and exit
-s, --show show image with label in title
-g, --gpu use gpu