yolo-streamlit-detection-tracking

YOLO Vision Studio: a free, open-source Streamlit app for detection, segmentation, pose, text prompts (YOLO World, YOLOE) and tracking with honest counts. Built

What is yolo-streamlit-detection-tracking?

YOLO Vision Studio: a free, open-source Streamlit app for detection, segmentation, pose, text prompts (YOLO World, YOLOE) and tracking with honest counts. Built on Ultralytics YOLO26.

Overview of yolo-streamlit-detection-tracking

What it is

YOLO Vision Studio is a free, open-source Streamlit application built on Ultralytics YOLO26. It performs object detection, segmentation, pose estimation, text-prompted detection (YOLO World v2, YOLOE) and object tracking with honest counting. The app shows both in-frame counts and total unique tracked objects, along with a churn metric that indicates tracker efficiency. It is developed by aparsoft and available on GitHub.

Key features

  • Supports detection, segmentation, pose estimation, YOLO World v2 (text prompts), and YOLOE (text to masks) on images and video.
  • Offers multiple model families: YOLO26 n/s/m/l/x, RT-DETR L/X, YOLO World v2 s/m/l/x, and YOLOE-26 n/s/m/l/x.
  • Accepts image uploads (JPG, PNG, BMP, WEBP) and video sources including stored files, webcam (via streamlit-webrtc), RTSP streams, and YouTube URLs.
  • Provides four trackers (ByteTrack, BoT-SORT, Deep OC-SORT, TrackTrack) with optional appearance ReID and tunable gates (IoU, appearance similarity, lost-track memory).
  • Includes speed controls: skip frames (1–8×), confidence slider, and class limiting. Display is capped at 30 fps for smooth browser performance.
  • Displays live metrics (frame number, FPS, in-frame counts, confirmed totals per class, churn) updated 5 times per second.
  • Supports multiple users: each browser session runs its own model instance, preventing tracker sharing.

Use cases

The app is designed for counting and tracking objects in video streams, such as pedestrians on a street, vehicles in traffic, or animals. Text prompts (e.g., “red car”) enable targeted detection of specific objects. Pose estimation is useful for analyzing body movements in activities like yoga. The honest counting mechanism avoids double-counting by requiring objects to be seen in at least 5 processed frames before being counted as unique. Sample videos are provided for testing different scenarios: people counting, mixed classes, vehicle tracking, pose estimation, and crossing paths.

Getting started

bash git clone https://github.com/aparsoft/yolo-streamlit-detection-tracking.git cd yolo-streamlit-detection-tracking uv sync # creates .venv with the locked versions uv run python scripts/get_sample_videos.py # optional: download sample clips (~34 MB) uv run streamlit run app.py # opens http://localhost:8501 Alternatively, use pip with a virtual environment. Python 3.10+ is required. Model weights download automatically on first use.

License and pricing

Licensed under Apache-2.0. The application is free and open-source.

Based on the project's public README and repository metadata.

Add this badge to your README

Featured on YunoDB Badge styles & help

Maintain this project? Show that it is featured on YunoDB.

Markdown[![Featured on YunoDB](https://yunodb.com/badge/aparsoftyolo-streamlit-detection-tracking.svg)](https://yunodb.com/tools/aparsoftyolo-streamlit-detection-tracking?utm_source=github&utm_medium=badge)
HTML<a href="https://yunodb.com/tools/aparsoftyolo-streamlit-detection-tracking?utm_source=github&utm_medium=badge"><img src="https://yunodb.com/badge/aparsoftyolo-streamlit-detection-tracking.svg" alt="Featured on YunoDB" height="20"></a>