Broken Pipeline for mask generation task
#10
by taher30 - opened
Using the model in Pipeline is way slower than using SamModel class. The pipeline takes 4 minutes to generate masks for a 400x500 image and given the same image, when used SamModel class it takes less than 40 seconds to process the image. The difference is significant.
I am using the below code for the Pipeline. Not sure what I am doing wrong here.
from transformers import pipeline
image = Image.open('sample.png')
pipe = pipeline("mask-generation", model="facebook/sam-vit-huge", device=0)
out=pipe(image)