Selective Search 실습과 IOU

computervision/섹션 1. Object Detection의 이해

Selective Search 실습과 IOU

Gbjbjbj 2021. 12. 25. 00:10

#ground truth의 좌표를 다음과 같이 설정함
import cv2
import matplotlib.pyplot as plt
%matplotlib inline

# 실제 box(Ground Truth)의 좌표를 아래와 같다고 가정. 
gt_box = [60, 15, 320, 420]


img = cv2.imread('./data/audrey01.jpg')
img_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)

red = (255, 0 , 0)
img_rgb = cv2.rectangle(img_rgb, (gt_box[0], gt_box[1]), (gt_box[2], gt_box[3]), color=red, thickness=2)

plt.figure(figsize=(8, 8))
plt.imshow(img_rgb)
plt.show()

##bounding box의 크기가 큰 후보만 추출

cand_rects = [cand['rect'] for cand in regions if cand['size'] > 10000]

green_rgb = (125, 255, 51)
img_rgb_copy = img_rgb.copy()
for rect in cand_rects:
    
    left = rect[0]
    top = rect[1]
    # rect[2], rect[3]은 너비와 높이이므로 우하단 좌표를 구하기 위해 좌상단 좌표에 각각을 더함. 
    right = left + rect[2]
    bottom = top + rect[3]
    
    img_rgb_copy = cv2.rectangle(img_rgb_copy, (left, top), (right, bottom), color=green_rgb, thickness=2)
    
plt.figure(figsize=(8, 8))
plt.imshow(img_rgb_copy)
plt.show()

!pip install selectivesearch #먼저 selectivesearch 모듈을 불러오자

import selectivesearch
import cv2
import matplotlib.pyplot as plt
import os
%matplotlib inline

### 오드리헵번 이미지를 cv2로 로드하고 matplotlib으로 시각화 
img = cv2.imread('./data/audrey01.jpg')
img_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
print('img shape:', img.shape)

plt.figure(figsize=(8, 8))
plt.imshow(img_rgb)
plt.show()

import selectivesearch 

#selectivesearch.selective_search()는 이미지의 Region Proposal정보를 반환 
#img_rgb에서 selectivesearch를 통해 찾은 bounding box정보를 반환
_, regions = selectivesearch.selective_search(img_rgb, scale=100, min_size=2000)

print(type(regions), len(regions))

<class 'list'> 41

반환된 regions 변수는 리스트 타입으로 세부 원소를 딕셔너리로 가지고 있음.

아래로 내려갈 수록 너비와 높이 값이 큰 Bounding box이며 하나의 Bounding box에 여러개의 오브젝트가

있을 확률이 커짐

ex) {'labels': [0.0], 'rect': (0, 0, 107, 167), 'size': 11166}

{'labels': [1.0], 'rect': (15, 0, 129, 110), 'size': 8771}

{'labels': [17.0, 18.0, 14.0, 16.0, 7.0, 11.0, 9.0, 12.0, 2.0, 6.0], 'rect': (0, 0, 374, 444), 'size': 61244}

cand_rects = [cand['rect'] for cand in regions]

# opencv의 rectangle()을 이용하여 시각화
# rectangle()은 이미지와 좌상단 좌표, 우하단 좌표, box컬러색, 두께등을 인자로 입력하면 원본 이미지에 box를 그려줌. 

green_rgb = (125, 255, 51)
img_rgb_copy = img_rgb.copy()
for rect in cand_rects:
    
    left = rect[0]
    top = rect[1]
    # rect[2], rect[3]은 너비와 높이이므로 우하단 좌표를 구하기 위해 좌상단 좌표에 각각을 더함. 
    right = left + rect[2]
    bottom = top + rect[3]
    #이미지에 rect형태로 표시
    img_rgb_copy = cv2.rectangle(img_rgb_copy, (left, top), (right, bottom), color=green_rgb, thickness=2)
    
plt.figure(figsize=(8, 8))
plt.imshow(img_rgb_copy)
plt.show()

#bounding box의 크기가 큰 후보만 추출한다

cand_rects = [cand['rect'] for cand in regions if cand['size'] > 10000]

green_rgb = (125, 255, 51)
img_rgb_copy = img_rgb.copy()
for rect in cand_rects:
    
    left = rect[0]
    top = rect[1]
    # rect[2], rect[3]은 너비와 높이이므로 우하단 좌표를 구하기 위해 좌상단 좌표에 각각을 더함. 
    right = left + rect[2]
    bottom = top + rect[3]
    
    img_rgb_copy = cv2.rectangle(img_rgb_copy, (left, top), (right, bottom), color=green_rgb, thickness=2)
    
plt.figure(figsize=(8, 8))
plt.imshow(img_rgb_copy)
plt.show()

IOU

Object detection 성능 평가 Metric - IOU(Intersection over Union)

모델이 예측한 결과와 실측 box가 얼마나 정확하게 겹치는가를 나타내는 지표

IOU = Area of overlap / Area of Union

코드로 살펴보자

import numpy as np 

def compute_iou(cand_box, gt_box):

    # Calculate intersection areas
    x1 = np.maximum(cand_box[0], gt_box[0])
    y1 = np.maximum(cand_box[1], gt_box[1])
    x2 = np.minimum(cand_box[2], gt_box[2])
    y2 = np.minimum(cand_box[3], gt_box[3])
    
    intersection = np.maximum(x2 - x1, 0) * np.maximum(y2 - y1, 0)
    
    cand_box_area = (cand_box[2] - cand_box[0]) * (cand_box[3] - cand_box[1])
    gt_box_area = (gt_box[2] - gt_box[0]) * (gt_box[3] - gt_box[1])
    union = cand_box_area + gt_box_area - intersection
    
    iou = intersection / union
    return iou

# 실제 box(Ground Truth)의 좌표를 아래와 같다고 가정한다
import cv2
import matplotlib.pyplot as plt
%matplotlib inline


gt_box = [60, 15, 320, 420]


img = cv2.imread('./data/audrey01.jpg')
img_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)

red = (255, 0 , 0)
img_rgb = cv2.rectangle(img_rgb, (gt_box[0], gt_box[1]), (gt_box[2], gt_box[3]), color=red, thickness=2)

plt.figure(figsize=(8, 8))
plt.imshow(img_rgb)
plt.show()

cand_rects = [cand['rect'] for cand in regions]
for index, cand_box in enumerate(cand_rects):
    cand_box = list(cand_box)
    cand_box[2] += cand_box[0]
    cand_box[3] += cand_box[1]
    
    iou = compute_iou(cand_box, gt_box)
    print('index:', index, "iou:", iou)

index: 0 iou: 0.06157293686705451

index: 1 iou: 0.07156308851224105

index: 2 iou: 0.2033654637255666

index: 3 iou: 0.04298195631528965

index: 4 iou: 0.14541310541310543

index: 5 iou: 0.10112060778727446

index: 6 iou: 0.11806905615946989

index: 7 iou: 0.1420163334272036

index: 8 iou: 0.035204259342190375

index: 9 iou: 0.004256894317971497..........

img = cv2.imread('./data/audrey01.jpg')
img_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
print('img shape:', img.shape)

green_rgb = (125, 255, 51)
#bound box의 크기가 3000이상인 것만 취급
cand_rects = [cand['rect'] for cand in regions if cand['size'] > 3000]
gt_box = [60, 15, 320, 420]
img_rgb = cv2.rectangle(img_rgb, (gt_box[0], gt_box[1]), (gt_box[2], gt_box[3]), color=red, thickness=2)

for index, cand_box in enumerate(cand_rects):
    
    cand_box = list(cand_box)
    cand_box[2] += cand_box[0]
    cand_box[3] += cand_box[1]
    
    iou = compute_iou(cand_box, gt_box)
    #iou가 0.5이상일 때만 화면에 표기
    if iou > 0.5:
        print('index:', index, "iou:", iou, 'rectangle:',(cand_box[0], cand_box[1], cand_box[2], cand_box[3]) )
        cv2.rectangle(img_rgb, (cand_box[0], cand_box[1]), (cand_box[2], cand_box[3]), color=green_rgb, thickness=1)
        text = "{}: {:.2f}".format(index, iou)
        cv2.putText(img_rgb, text, (cand_box[0]+ 100, cand_box[1]+10), cv2.FONT_HERSHEY_SIMPLEX, 0.4, color=green_rgb, thickness=1)
    
plt.figure(figsize=(12, 12))
plt.imshow(img_rgb)
plt.show()