facebookresearch / facebookresearch/segment-anything

set_image/image embedding runtime

Open
#229 2 comments 0 reactions 0 assignees View on GitHub
Dominant language
Jupyter Notebook
Stars
54.9k
Forks
6.4k
PR merge metrics
No merged PRs in 30d

Description

Hello, I was wondering what sort of run-times you guys are getting for the set_image function. I am using an RTX 2060 and running the vit_l model, averaging ~1.3s for just the image_set function and ~0.3s for the actual NN processing. These numbers seem to be resolution independent. I was curious to see if you guys are getting varying performance for different GPUs and vit_model/checkpoint used, please state both if possible.

![image](https://user-images.githubusercontent.com/130655864/232413477-a1109457-4511-46cc-89d3-68d11b47ab49.png)

Additionally, has anyone managed to lower the image_set time taken?

Contributor guide

Open the contributing guide

Research direction

No file, test, or entry point is named. Start by reproducing the set_image/image_set timing with the vit_l model on the reported RTX 2060, then compare it with the reported neural-network processing time. Done would require identifying whether the overhead varies by GPU or checkpoint and documenting a supported improvement or explanation.

Written by the indexing model from the issue text.

Assessment

Tech stack
jupyter-notebook
Domain
computer-vision, machine-learning, performance
Issue type
Bug
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
20/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.