hyun5 / as-one

Easy & Modular Computer Vision Detectors and Trackers - Run YOLOv8,v7,v6,v5,R,X in under 20 lines of code.

AS-One : A Modular Libary for YOLO Object Detection and Object Tracking BETA

Table of Contents

1. Introduction
2. Prerequisites
3. Clone the Repo
4. Installation
   • Linux
   • Windows 10/11
5. Running AS-One
6. Sample Code Snippets
7. Benchmarks

1. Introduction

==UPDATE: YOLOv8 Now Supported==

AS-One is a python wrapper for multiple detection and tracking algorithms all at one place. Different trackers such as ByteTrack, DeepSort or NorFair can be integrated with different versions of YOLO with minimum lines of code. This python wrapper provides YOLO models in both ONNX and PyTorch versions. We plan to offer support for future versions of YOLO when they get released.

This is One Library for most of your computer vision needs.

If you would like to dive deeper into YOLO Object Detection and Tracking, then check out our courses and projects

Watch the step-by-step tutorial

2. Prerequisites

• Make sure to install GPU drivers in your system if you want to use GPU . Follow driver installation for further instructions.
• Make sure you have MS Build tools installed in system if using windows.
• Download git for windows if not installed.

3. Clone the Repo

Navigate to an empty folder of your choice.

git clone https://github.com/augmentedstartups/AS-One.git

Change Directory to AS-One

cd AS-One

4. Installation

For Linux

python3 -m venv .env
source .env/bin/activate

pip install numpy Cython
pip install cython-bbox

pip install asone


# for CPU
pip install torch torchvision

# for GPU
pip install torch torchvision --extra-index-url https://download.pytorch.org/whl/cu113

For Windows 10/11

python -m venv .env
.env\Scripts\activate
pip install numpy Cython
pip install -e git+https://github.com/samson-wang/cython_bbox.git#egg=cython-bbox

pip install asone

# for CPU
pip install torch torchvision

# for GPU
pip install torch torchvision --extra-index-url https://download.pytorch.org/whl/cu113
or
pip install torch==1.10.1+cu113 torchvision==0.11.2+cu113 torchaudio===0.10.1+cu113 -f https://download.pytorch.org/whl/cu113/torch_stable.html

5. Running AS-One

Run main.py to test tracker on data/sample_videos/test.mp4 video

python main.py data/sample_videos/test.mp4

Run in Google Colab

6. Sample Code Snippets

6.1. Object Detection

import asone
from asone import utils
from asone import ASOne
import cv2

video_path = 'data/sample_videos/test.mp4'
detector = ASOne(detector=asone.YOLOV7_PYTORCH, use_cuda=True) # Set use_cuda to False for cpu

filter_classes = ['car'] # Set to None to detect all classes

cap = cv2.VideoCapture(video_path)

while True:
    _, frame = cap.read()
    if not _:
        break

    dets, img_info = detector.detect(frame, filter_classes=filter_classes)

    bbox_xyxy = dets[:, :4]
    scores = dets[:, 4]
    class_ids = dets[:, 5]

    frame = utils.draw_boxes(frame, bbox_xyxy, class_ids=class_ids)

    cv2.imshow('result', frame)

    if cv2.waitKey(25) & 0xFF == ord('q'):
        break

Run the asone/demo_detector.py to test detector.

# run on gpu
python -m asone.demo_detector data/sample_videos/test.mp4

# run on cpu
python -m asone.demo_detector data/sample_videos/test.mp4 --cpu

6.1.1 Use Custom Trained Weights for Detector

Use your custom weights of a detector model trained on custom data by simply providing path of the weights file.

import asone
from asone import utils
from asone import ASOne
import cv2

video_path = 'data/sample_videos/license_video.webm'
detector = ASOne(detector=asone.YOLOV7_PYTORCH, weights='data/custom_weights/yolov7_custom.pt', use_cuda=True) # Set use_cuda to False for cpu

class_names = ['license_plate'] # your custom classes list

cap = cv2.VideoCapture(video_path)

while True:
    _, frame = cap.read()
    if not _:
        break

    dets, img_info = detector.detect(frame)

    bbox_xyxy = dets[:, :4]
    scores = dets[:, 4]
    class_ids = dets[:, 5]

    frame = utils.draw_boxes(frame, bbox_xyxy, class_ids=class_ids, class_names=class_names) # simply pass custom classes list to write your classes on result video

    cv2.imshow('result', frame)

    if cv2.waitKey(25) & 0xFF == ord('q'):
        break

6.1.2. Changing Detector Models

Change detector by simply changing detector flag. The flags are provided in benchmark tables.

# Change detector
detector = ASOne(detector=asone.YOLOX_S_PYTORCH, use_cuda=True)

6.2. Object Tracking

Use tracker on sample video.

import asone
from asone import ASOne

# Instantiate Asone object
dt_obj = ASOne(tracker=asone.BYTETRACK, detector=asone.YOLOV7_PYTORCH, use_cuda=True) #set use_cuda=False to use cpu

filter_classes = ['person'] # set to None to track all classes

# ##############################################
#           To track using video file
# ##############################################
# Get tracking function
track_fn = dt_obj.track_video('data/sample_videos/test.mp4', output_dir='data/results', save_result=True, display=True, filter_classes=filter_classes)

# Loop over track_fn to retrieve outputs of each frame 
for bbox_details, frame_details in track_fn:
    bbox_xyxy, ids, scores, class_ids = bbox_details
    frame, frame_num, fps = frame_details
    # Do anything with bboxes here

# ##############################################
#           To track using webcam
# ##############################################
# Get tracking function
track_fn = dt_obj.track_webcam(cam_id=0, output_dir='data/results', save_result=True, display=True, filter_classes=filter_classes)

# Loop over track_fn to retrieve outputs of each frame 
for bbox_details, frame_details in track_fn:
    bbox_xyxy, ids, scores, class_ids = bbox_details
    frame, frame_num, fps = frame_details
    # Do anything with bboxes here

# ##############################################
#           To track using web stream
# ##############################################
# Get tracking function
stream_url = 'rtsp://wowzaec2demo.streamlock.net/vod/mp4:BigBuckBunny_115k.mp4'
track_fn = dt_obj.track_stream(stream_url, output_dir='data/results', save_result=True, display=True, filter_classes=filter_classes)

# Loop over track_fn to retrieve outputs of each frame 
for bbox_details, frame_details in track_fn:
    bbox_xyxy, ids, scores, class_ids = bbox_details
    frame, frame_num, fps = frame_details
    # Do anything with bboxes here

[Note] Use can use custom weights for a detector model by simply providing path of the weights file. in ASOne class.

6.2.1 Changing Detector and Tracking Models

Change Tracker by simply changing the tracker flag.

The flags are provided in benchmark tables.

dt_obj = ASOne(tracker=asone.BYTETRACK, detector=asone.YOLOV7_PYTORCH, use_cuda=True)
# Change tracker
dt_obj = ASOne(tracker=asone.DEEPSORT, detector=asone.YOLOV7_PYTORCH, use_cuda=True)

# Change Detector
dt_obj = ASOne(tracker=asone.DEEPSORT, detector=asone.YOLOX_S_PYTORCH, use_cuda=True)

To setup ASOne using Docker follow instructions given in docker setup

ToDo

• First Release
• Import trained models
• Simplify code even further
• Updated for YOLOv8
• OCSORT, StrongSORT, MoTPy
• OCR and Counting
• M1/2 Apple Silicon Compatibility

Offered By:	Maintained By: