Practical overview
When it fits
Use As One when developing in Python and you want a single library interface for selecting supported YOLO detectors and trackers for image or video workflows.
Who it is for
- Python developers building computer-vision applications or prototypes.
- Teams processing video that need object detection, tracking and annotated outputs.
- Developers experimenting with different YOLO detector and tracker combinations.
Documented uses
- Run object detection over video frames.
- Track detected objects in video and filter results by class.
- Draw annotations from detection or tracking outputs.
- Use custom-trained detector weights.
- Build segmentation or pose-estimation workflows using the library's stated model integrations.
Requirements & limits
The project is published as an open-source repository under GPL-3.0. The README documents pip installation and source installation; Windows users are directed to install from source.
- The README lists ONNX, PyTorch and CoreML model formats.
- Installation steps differ by operating system; Windows has a source-install requirement in the README.
- GPU use requires GPU drivers, while examples also describe CPU operation.
- The documented video-tracking example uses a chosen detector and tracker and yields model outputs that can be drawn as annotations.
- The repository identifies Python as its primary language and GPL-3.0 as its license.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Track selected objects in a video
Use the documented sample pattern to run a detector and tracker on a video and draw annotations.
Before you start
- Python environment with As One and its dependencies installed.
- A video file accessible at the path used in the example.
- GPU drivers if using GPU acceleration; the example sets use_cuda=True.
Steps
Import asone and ASOne.
Create an ASOne instance with the BYTETRACK tracker and YOLOV9_C detector.
Call video_tracker on the video path, optionally filtering for a class such as car.
Iterate through outputs and call ASOne.draw for each model output.
Expected outcome
A sequence of tracking outputs for the selected video and class, with annotations generated for each output.
Guide limitations
- The README's example uses a sample video path; provide a video available in your environment.
- The repository gives CPU/GPU setup directions, but this example specifically sets use_cuda=True.
Install As One from source on Linux
Prepare the documented Linux source environment for using the library.
Before you start
- Git and Python 3.
- An empty folder for cloning the repository.
- GPU drivers only if intending to use GPU.
Steps
Clone the AS-One repository and change into the AS-One directory.
Create and activate a Python virtual environment.
Install the repository requirements.
Install torch and torchvision; for GPU, use the documented CUDA 11.3 package index option.
Expected outcome
A Linux virtual environment with repository requirements and the selected CPU or GPU PyTorch packages installed.
Guide limitations
- The documented steps cover installation, not verification or running a project-specific workflow.
- GPU driver installation is a separate prerequisite when using GPU.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 15 October 2026
- Original website title
- augmentedstartups/AS-One · Official homepage title · retained in its source language