Is this the correct usage for stream tracking?

[tonyreina]

This code seems to be working correctly, but I am a little unsure of my using the for result loop and was hoping to get some feedback.

From what I can tell, the model.track when set as a stream=True will work on batches of images within the given directory path. However, it looks like it won't actually perform the function until the iterator is called. So I'm just running a pass in the iterator for-loop. This would then save the tracking info for each frame in a .txt file with the frame number.

Is that the correct usage? Is there a more elegant way to do this?

Thanks.
Best.
-Tony

import os
from ultralytics import YOLO
from tqdm import tqdm
from glob import glob

class YoloTracking:
    """Track cells in frames using YOLO."""

    def __init__(self, cfg):
        self.weights_pth = cfg.track.weights
        self.device = cfg.track.device
        self.confidence = cfg.track.conf
        self.iou = cfg.track.iou
        self.max_det = cfg.track.max_det
        self.source_folder = cfg.expt.paths.bg_out
        self.destination_folder = cfg.expt.paths.cell_kinetics
        self.input_image = cfg.expt.images.output
        self.experiment_label = f"{cfg.expt.project.benchling_id}_{cfg.expt.project.run_id}"
        self.image_max_dim = getattr(cfg.track, "image_max_dim", 3264)
        self.batch_size = getattr(cfg.track, "batch_size", 32)

        self.model = YOLO(self.weights_pth, task="detect")

    def process(self):
        """Process images using YOLO tracking."""
        try:
            # Ensure the destination folder exists
            os.makedirs(self.destination_folder, exist_ok=True)

            # Pattern for input files
            pattern = os.path.join(self.source_folder, f"*{self.input_image}")

            frames = glob(pattern)
            
            # Run YOLO tracking
            results = self.model.track(
                pattern,
                persist=True,
                imgsz=self.image_max_dim,
                conf=self.confidence,
                iou=self.iou,
                max_det=self.max_det,
                device=self.device,
                stream=True,
                save_txt=True,
                half=True,
                batch=self.batch_size,
                verbose=False,
                project=self.destination_folder,
                name=self.experiment_label
            )

            for result in tqdm(results, total=len(frames)):
                pass

            print("Tracking completed successfully.")

        except Exception as e:
            print(f"An error occurred during processing: {e}")

[UltralyticsAssistant]

👋 Hello @tonyreina, thank you for your interest in Ultralytics 🚀! Your use case sounds intriguing—thanks for sharing your detailed code and explanation. An Ultralytics engineer will review this soon and provide further insights 🎯.

In the meantime, here are a few suggestions and resources that may help:

---

General Guidance

• If this is a ❓ Question, please provide additional details, such as any specific challenges you're encountering or what outputs you're observing. This will help us provide more targeted feedback.
• If this is a 🐛 Bug Report, please consider including a minimum reproducible example to make it easier to debug the behavior you're describing.

---

Upgrade

Ensure you’re using the latest ultralytics package to avoid encountering issues that may already have been resolved. Upgrade in a Python>=3.8 environment with PyTorch>=1.8:

pip install -U ultralytics

---

Explore the Docs 📚

Our documentation includes a wealth of information about supported features and usage examples:

• Python Usage Examples
• Tracking with YOLO

In particular, the streaming and tracking section might help clarify how the model.track() function operates when the stream=True argument is applied and might offer alternative approaches to your loop implementation.

---

Join Our Community 🧑‍🤝‍🧑

Have you checked out our Ultralytics community? Here's where you can connect with other users and find additional support:

• Discord for real-time discussions 🎧
• Discourse for in-depth conversations 📝
• Subreddit to share and discuss your projects 🚀

---

Environments

In case you want to test in different setups, YOLO works in several verified environments, including free GPU Notebooks:

• 
• 
• 

For deployment scenarios, see Google Cloud, AWS, or our Docker guide here.

---

Status

Alternatively, to ensure everything is functioning smoothly with your environment setup, check the Ultralytics CI badge status below. CI tests verify functionality across various platforms, tasks, and modes every 24 hours and on each commit:

Let us know if you have additional concerns—we’re here to help! 😊

[glenn-jocher]

@tonyreina your usage of model.track with stream=True is correct. The iterator (for result in results) triggers processing, and the pass statement is appropriate if you don't need to handle results in-memory. This setup efficiently saves tracking data directly to .txt files. For a more elegant approach, you could process or log specific results within the loop if needed. Let us know if you need further clarification!