Prediction on SAM Model doesn't support specifying classes to predict about ultralytics HOT 4 CLOSED

👋 Hello @zoltanmaric, thank you for your interest in Ultralytics YOLOv8 🚀! We recommend a visit to the Docs for new users where you can find many Python and CLI usage examples and where many of the most common questions may already be answered.

If this is a 🐛 Bug Report, please provide a minimum reproducible example to help us debug it.

If this is a custom training ❓ Question, please provide as much information as possible, including dataset image examples and training logs, and verify you are following our Tips for Best Training Results.

Join the vibrant Ultralytics Discord 🎧 community for real-time conversations and collaborations. This platform offers a perfect space to inquire, showcase your work, and connect with fellow Ultralytics users.

Install

Pip install the ultralytics package including all requirements in a Python>=3.8 environment with PyTorch>=1.8.

pip install ultralytics

Environments

YOLOv8 may be run in any of the following up-to-date verified environments (with all dependencies including CUDA/CUDNN, Python and PyTorch preinstalled):

• Notebooks with free GPU:
• Google Cloud Deep Learning VM. See GCP Quickstart Guide
• Amazon Deep Learning AMI. See AWS Quickstart Guide
• Docker Image. See Docker Quickstart Guide

Status

If this badge is green, all Ultralytics CI tests are currently passing. CI tests verify correct operation of all YOLOv8 Modes and Tasks on macOS, Windows, and Ubuntu every 24 hours and on every commit.

from ultralytics.

@zoltanmaric hello!

Thank you for reaching out with your query. Currently, the SAM model in YOLOv8 doesn't directly support the classes parameter for limiting the range of classes detected during predictions. The behavior you observed is linked to how the SAM model is configured to handle predictions broadly, without filtering by specific classes post-training.

However, a potential workaround is to manually filter the output results after predictions are made. Here's a quick example of how you might do this:

from ultralytics import SAM

model = SAM('sam_b.pt')
results = model.predict('path/to/image.jpg')
filtered_results = [r for r in results if r['class'] in (1, 3, 6)]

This approach involves filtering the results to include only the desired classes after the prediction has been run. While this isn't as efficient as direct class filtering in the predict function, it provides a method to achieve your goal.

We acknowledge the need for this feature and will consider it for future updates. If you have any more suggestions or need further assistance, feel free to ask.

Best regards!

from ultralytics.

Thanks @glenn-jocher.

Yeah, filtering the results after-the-fact is what I also did, but I noticed the SAM model is quite slow at inference (taking 2-5 minutes on a MacBook Air with an M1 processor), so I was hoping to speed it up some by restricting the number of classes it's looking for. Thanks anyway!

from ultralytics.

Hi @zoltanmaric,

Thanks for your follow-up! Indeed, filtering post-prediction isn't the most efficient for speed, especially with the SAM model's comprehensive segmentation capabilities. Currently, there isn't a built-in feature to limit classes pre-inference in the SAM model, which could indeed help speed things up as you mentioned.

As a potential alternative, you might consider using a more lightweight model if your use case permits, or exploring hardware acceleration options that could handle the SAM model's demands more effectively.

We appreciate your input and will keep it in mind for future updates to enhance performance and usability. Thanks for using Ultralytics YOLO! 🚀

Best regards!

from ultralytics.