환경설정(colab)

from google.colab import drive
drive.mount('/content/gdrive')

폴더 경로 설정

workspace_path = '/content/gdrive/경로'

필요 패키지 로드

import cv2
import os
import torch
import random
import numpy as np
import torch.nn as nn
import torch.optim as optim
import matplotlib.pyplot as plt
import torch.nn.functional as F
from torchvision import datasets, transforms
from PIL import Image

Object Detection (YOLOv3)

YOLOv3 코드 다운로드

!git clone https://github.com/AlexeyAB/darknet.git

Colab에서 Darknet의 GPU 지원 활성화

%cd darknet
!sed -i 's/GPU=0/GPU=1/' Makefile
!sed -i 's/CUDNN=0/CUDNN=1/' Makefile
!sed -i 's/OPENCV=0/OPENCV=1/' Makefile
!make

YOLOv3 사전 훈련된 가중치 다운로드

!wget https://pjreddie.com/media/files/yolov3.weights

dog.jpg에 대해 인퍼런스 수행

!./darknet detect cfg/yolov3.cfg yolov3.weights data/dog.jpg

Object Detection (YOLOv8)

%cd /content
!pip install ultralytics

from ultralytics import YOLO 
import io  ## https://github.com/ultralytics/ultralytics
from urllib import request

test_url = 'https://mediahub.seoul.go.kr/wp-content/uploads/2016/06/4ff235248c9e8c458dec3c0c42b29146.jpg'
img = Image.open(io.BytesIO(request.urlopen(test_url).read()))
plt.imshow(img)
plt.show()

model = YOLO('yolov8m.pt')  
results = model.predict(source=test_url) 
#results = model.track(source=test_url)  ## Tracking

for r in results :
  box = r.boxes
  im_array = r.plot()  # plot a BGR numpy array of predictions
  im = Image.fromarray(im_array[..., ::-1])  # RGB PIL image
  im.save('results.jpg')

predict_img = Image.open('results.jpg')
plt.imshow(predict_img)
plt.show()

#결과:

• YOLOv8은 Object Detection 외에도 Classification, Instance Segmentation, Pose Estimation, Tracking을 지원함

YOLOv8 Seg

model = YOLO('yolov8x-seg.pt')
results = model.predict(source=test_url)

for r in results :
  im_array = r.plot()  # plot a BGR numpy array of predictions
  im = Image.fromarray(im_array[..., ::-1])  # RGB PIL image
  im.save('seg_results.jpg')

predict_img = Image.open('seg_results.jpg')
plt.imshow(predict_img)
plt.show()

#결과 :

Object Tracking (Bytetrack)

바이트 트랙 코드 다운로드

# == Download the repo content and install dependencies ==
%cd /content
!git clone https://github.com/ifzhang/ByteTrack.git
%cd /content/ByteTrack/
%mkdir pretrained
%cd pretrained

# == Download pretrained X model weights ==
!gdown --id "1P4mY0Yyd3PPTybgZkjMYhFri88nTmJX5"

COCO API 다운, 패키지 설치

# == Install dependencies ==
!pip3 install cython
!pip3 install 'git+https://github.com/cocodataset/cocoapi.git#subdirectory=PythonAPI'
!pip3 install cython_bbox
!pip3 install loguru
!pip3 install thop
!pip3 install lap

%cd /content/ByteTrack/
!pip3 install -r requirements.txt

# == Install ByteTrack ==
!python3 setup.py develop

트래킹 코드 실행

# run inference demo (can be slow on colab). The cell output is deflected to the file 'log.txt' for downstream use to display the result but feel free to remove
%cd /content/ByteTrack
!python3 tools/demo_track.py video -f exps/example/mot/yolox_x_mix_det.py -c pretrained/bytetrack_x_mot17.pth.tar --fp16 --fuse --save_result &> log.txt

# == Get rendered result video file path ==
import re
%cd /content/ByteTrack
with open('log.txt', 'r') as file:
    text = file.read().replace('\n', '')

m = re.search('video save_path is ./(.+?).mp4', text) #mp4 형식의 파일을 찾아냄
print(m)
if m:
    found = '/content/ByteTrack/' + m.group(1) + ".mp4"

found

Colab에서 영상 재생을 위해 MP4 -> webm 파일로 변환

# == Convert mp4 to webm video file for proper display on colab ==
%cd /content/ByteTrack
!ffmpeg -i $found -vcodec vp9 out.webm

webm 파일을 재생하는 코드

# == Display results ==
from IPython.display import HTML
from base64 import b64encode

video_path = '/content/ByteTrack/out.webm'

mp4 = open(video_path, "rb").read()
data_url = "data:video/webm;base64," + b64encode(mp4).decode()
HTML(f"""
<video width=600 controls>
      <source src="{data_url}" type="video/webm">
</video>
""")

#결과: