Incredible, Segment Anything acceleration version

Traditional SAM is great but a little slow

SAM (Segment Anything) is a model that allows you to segment any object in any image into instances .
You can segment anything by specifying points.

SegmentAnything is amazing because it can cleanly segment anything (cut out objects)

However, previous models took about 1 second to capture each image.

Here’s a really fast model

FastSAM is a version that is about 10 times faster .
This opens up the possibility of using it for various tasks such as real-time segmentation.

GitHub - CASIA-IVA-Lab/FastSAM: Fast Segment Anything

What’s more, you can freely cut out objects specified by text, boxes, or points.

Easy to use

And it’s easy to use. Below are the steps.

install

git clone https://github.com/CASIA-IVA-Lab/FastSAM.git # clone Fast-SAM
cd FastSAM
pip install -r requirements.txt
pip install git+https://github.com/openai/CLIP.git # install CLIP

Segment all instances

python Inference.py --model_path FastSAM-x.pt --img_path ./images/students.jpg

black hair girl

python Inference.py --model_path FastSAM-x.pt --img_path ./images/students.jpg  --text_prompt "the black hair girl"

Segment object specified by box coordinates

python Inference.py --model_path FastSAM-x.pt --img_path ./images/students.jpg --box_prompt "[[290,200,300,500]]"

Segment object specified by points

python Inference.py --model_path FastSAM-x.pt --img_path ./images/students.jpg --point_prompt "[[600,360],[900,360]]" --point_label "[1,1]"

Use from Python code

from fastsam import FastSAM, FastSAMPrompt

model = FastSAM('FastSAM-x.pt')
IMAGE_PATH = './images/girls.jpg'
DEVICE = "cuda:0"

everything_results = model(IMAGE_PATH, device=DEVICE, retina_masks=True, imgsz=1024, conf=0.4, iou=0.9,)
prompt_process = FastSAMPrompt(IMAGE_PATH, everything_results, device=DEVICE)

# everything prompt
# ann = prompt_process.everything_prompt()

# bbox default shape [0,0,0,0] -> [x1,y1,x2,y2]
# ann = prompt_process.box_prompt(bbox=[100, 200, 300, 700])

# text prompt
ann = prompt_process.text_prompt(text='skirt')

# point prompt
# points default [[0,0]] [[x1,y1],[x2,y2]]
# point_label default [0] [1,0] 0:background, 1:foreground
# ann = prompt_process.point_prompt(points=[[200, 360]], pointlabel=[1])

prompt_process.plot(annotations=ann,output_path='./output/students.jpg',)

That’s amazing.

🐣

I’m a freelance engineer.
Work consultation
Please feel free to contact us with a brief development description.
rockyshikoku@gmail.com

I am creating applications using machine learning and AR technology.

I send machine learning / AR related information.

GitHub

Twitter
Medium