1 Introduction
2 Evolutionary and swarm algorithms
2.1 Evolutionary algorithms
GA | DE | |
---|---|---|
Parents selection | Fitness-based selection | Random selection |
Crossover | Genes swapping between two parents to generate two offspring | Components selection from parent (target vector) and donor vector to generate one offspring (trial vector) |
Mutation | Bit inversion of each gene | Generation of one donor vector from three individuals |
Survivors selection | Complete replacement | Fitness-based competition |
2.2 Swarm algorithms
PSO | ACO | |
---|---|---|
Information shared within the swarm | Global best position (gbest) | Pheromone intensity |
Information that affects agent movement | Best positions for global (gbest) and personal (pbest) | Pheromone intensity and heuristic information |
2.3 Comparison of algorithm characteristics
3 Applications in computer vision
4 Neural network
4.1 Evolving DNNs for image classification
-
LEIC
-
EvoCNN
-
CoDeepNEAT
-
CGP-CNN
-
Genetic CNN
-
HREAS
-
DENSER
-
PSOAO
-
IPPSO
Dataset | Approach | Accuracy | OPT. time | #GPUs |
---|---|---|---|---|
CIFAR-10 | LEIC [1] | 94.60 | 10 days | 250 |
CoDeepNEAT [3] | 92.70 | – | – | |
CGP-CNN [4] | 94.34 | 10.4 days | 2 | |
Genetic CNN [5] | 92.90 | 2 days | 10 | |
HREAS [6] | 96.40 | 1.5 days | 200 | |
93.29 | – | – | ||
PSOAO [10] | 83.5 | 3.4 days | 1 | |
CIFAR-100 | LEIC [1] | 77.00 | – | – |
Genetic CNN [5] | 70.97 | – | – | |
77.51 | – | – | ||
MNIST | EvoCNN [2] | 98.82 | 2-3 days | 2 |
99.70 | – | – | ||
PSOAO [10] | 99.51 | 5 days | 1 | |
IPPSO [11] | 98.87 | – | – | |
BZI000065_conv-xml0x.pngFashion MNIST\c:IMG | EvoCNN [2] | 94.53 | 4 days | 2 |
95.11 | – | – | ||
ImageNet | Genetic CNN [5] | 72.13 | – | – |
HREAS [6] | 79.70 | – | – |
4.2 Evolving DNNs for image restoration
-
DPPN
-
E-CAE
4.3 Evolving DNNs for other tasks
Algorithm | Author | Section |
---|---|---|
GA | Real et al. [1] | Section 4.1 |
Sun et al. [2] | Section 4.1 | |
Miikkulainen et al. [3] | Section 4.1 | |
Suganuma et al. [4] | Section 4.1 | |
Xie and Yuille [5] | Section 4.1 | |
Liu et al. [6] | Section 4.1 | |
Section 4.1 | ||
Kramer [9] | Section 4.1 | |
Fernando et al. [12] | Section 4.2 | |
Suganuma et al. [13] | Section 4.2 | |
Zhining and Yunming [15] | Section 4.3 | |
Ouellette et al. [14] | Section 4.3 | |
PSO | Sun et al. [10] | Section 4.1 |
Wang et al. [11] | Section 4.1 |
5 Image segmentation
5.1 Thresholding approaches
-
Fuzzy partition×
-
Otsu Method
5.2 Clustering approaches
5.3 Other approaches
Algorithm | Author | Section |
---|---|---|
GA | Tao et al. [16] | Section 5.1 |
Maulik and Bandyopadhyay [26] | Section 5.2 | |
Awad et al. [28] | Section 5.2 | |
Awad et al. [29] | Section 5.2 | |
Halder et al. [31] | Section 5.2 | |
Halder et al. [32] | Section 5.2 | |
Pignalberi et al. [34] | Section 5.3 | |
Jiang et al. [35] | Section 5.3 | |
PSO | Puranik et al. [18] | Section 5.1 |
Ghamisi et al. [20] | Section 5.1 | |
Chander et al. [23] | Section 5.1 | |
Omran et al. [24] | Section 5.2 | |
Omran et al. [27] | Section 5.2 | |
ACO | Tao et al. [17] | Section 5.1 |
Liang et al. [19] | Section 5.1 | |
Liang and Yin [21] | Section 5.1 | |
Liang and Yin [22] | Section 5.1 | |
Malisia and Tizhoosh [25] | Section 5.2 | |
Bansal and Aggarwal [30] | Section 5.2 | |
Ouadfel and Batouche [33] | Section 5.3 | |
Wang et al. [36] | Section 5.3 | |
Ma et al. [37] | Section 5.3 |
6 Feature detection and selection
6.1 Feature detection
6.2 Feature selection
Algorithm | Author | Section |
---|---|---|
GA | Dong et al. [41] | Section 6.1 |
Trujillo and Olague [42] | Section 6.1 | |
Trujillo and Olague [43] | Section 6.1 | |
Perez and Olague [44] | Section 6.1 | |
Perez and Olague [45] | Section 6.1 | |
Yu et al. [46] | Section 6.2 | |
Treptow and Zell [47] | Section 6.2 | |
DE | Cuevas et al. [40] | Section 6.1 |
Khushaba et al. [48] | Section 6.2 | |
Khushaba et al. [49] | Section 6.2 | |
Gosh et al. [50] | Section 6.2 | |
PSO | Dong et al. [41] | Section 6.1 |
Ghamisi et al. [51] | Section 6.2 | |
Ghamisi et al. [52] | Section 6.2 | |
ACO | Nezamabadi-pour et al. [38] | Section 6.1 |
Baterina and Oppus [39] | Section 6.1 | |
Al-Ani [53] | Section 6.2 | |
Chen et al. [54] | Section 6.2 |
7 Image matching
7.1 Template matching
7.2 Image registration
7.3 Jigsaw-puzzle-like problems
Approach | Type | Number of pieces | Average best |
---|---|---|---|
Sholomon et al. [65] | Type 1 | 22,834 | 96.28 |
Sholomon et al. [66] | 30,745 | 93.40 | |
Sholomon et al. [67] | Type 2 | 22,755 | 91.07 |
Type 4 | 10,375 | 99.20 |
7.4 Feature matching
Algorithm | Author | Section |
---|---|---|
GA | Zhang and Akashi [55] | Section 7.1 |
Zhang and Akashi [56] | Section 7.1 | |
Zhang and Akashi [57] | Section 7.1 | |
Sato and Akashi [58] | Section 7.1 | |
Lee et al. [59] | Section 7.1 | |
Sholomon et al. [65] | Section 7.3 | |
Sholomon et al. [66] | Section 7.3 | |
Sholomon et al. [67] | Section 7.3 | |
Sizikova and Funkhouser [114] | Section 7.3 | |
Myers and Hancock [68] | Section 7.4 | |
Zhang et al. [69] | Section 7.4 | |
DE | Ugolotti et al. [60] | Section 7.1 |
De Falco et al. [61] | Section 7.2 | |
Ma et al. [62] | Section 7.2 | |
PSO | Ugolotti et al. [60] | Section 7.1 |
Wachowiak et al. [63] | Section 7.2 | |
Liebelt and Schertler [64] | Section 7.2 |
8 Visual tracking
8.1 Single object tracking
8.2 Multiple object tracking
Algorithm | Author | Section |
---|---|---|
GA | Bhaskar et al. [70] | Section 8.1 |
Huang and Essa [76] | Section 8.2 | |
DE | Cuevas et al. [71] | Section 8.1 |
Lin and Zhu [74] | Section 8.1 | |
Nenavath and Jatoth [75] | Section 8.1 | |
PSO | Zhang et al. [72] | Section 8.1 |
Cheng et al. [73] | Section 8.1 | |
Zhang et al. [77] | Section 8.2 |
9 Face recognition
9.1 Feature selection/weighting
-
Feature selection
-
Feature weighting
Database | Approach | #Individuals | #Training images per individual | #Test images per individual | Recognition rate |
---|---|---|---|---|---|
FERET database | Liu and Wechsler [78] | 369 | 2 | 1 | 92.14 |
Zheng et al. [79] | 255 | 3 | 1 | 89.37 | |
Senaratne et al. [83] | 1195 | 1 | 1 | 97.1 | |
CMU PIE database | Zheng et al. [79] | 68 | 3 | 18 | 96.81 |
Essex face database | Vignolo et al. [80] | 100 | 5 | 15 | 98.00 |
ORL database | Kanan et al. [81] | 40 | - | - | 99.75 |
Ramadan and Abdel-Kader [82] | 40 | 4 | 6 | 96.8 | |
Plastic surgery face database | Bhatt et al. [84] | 540 | - | - | 87.32 |
Combined heterogeneous face database | Bhatt et al. [84] | 1080 | - | - | 89.87 |
9.2 Fusion of visible and IR features
9.3 Other methods
Algorithm | Author | Section |
---|---|---|
GA | Liu and Wechsler [78] | Section 9.1 |
Zheng et al. [79] | Section 9.1 | |
Vignolo et al. [80] | Section 9.1 | |
Bhatt et al. [84] | Section 9.1 | |
Bebis et al. [85] | Section 9.2 | |
Desa and Hati [86] | Section 9.2 | |
Hermosilla et al. [87] | Section 9.2 | |
Wong et al. [88] | Section 9.3 | |
Akashi et al. [89] | Section 9.3 | |
Sato and Akashi [93] | Section 9.3 | |
You and Akashi [94] | Section 9.3 | |
DE | Chandar and Savithri [92] | Section 9.3 |
PSO | Ramadan and Abdel-Kader [82] | Section 9.1 |
Senaratne et al. [83] | Section 9.1 | |
Perez et al. [90] | Section 9.3 | |
Mpiperis et al. [91] | Section 9.3 | |
ACO | Kanan et al. [81] | Section 9.1 |
Mpiperis et al. [91] | Section 9.3 |
10 Human action recognition
10.1 Human body
Dataset | Approach | Evaluation index | Performance |
---|---|---|---|
MSR-Action3D dataset | Chaaraoui and Flórez-Revuelta [98] | Recognition rate | 93.10 |
Chaaraoui et al. [99] | 93.23 | ||
UCF50 dataset | Ijjina and Chalavadi [100] | Recognition rate | 99.98 |
CAD-60 | Nunes et al. [101] | Precision | 81.83 |
Recall | 80.02 |
10.2 Human hands
10.3 Human head
Algorithm | Author | Section |
---|---|---|
GA | Chaaraoui and Flórez-Revuelta [98] | Section 10.1 |
Chaaraoui et al. [99] | Section 10.1 | |
Ijjina and Chalavadi [100] | Section 10.1 | |
Oikonomidis et al. [105] | ||
DE | Nunes et al. [101] | Section 10.1 |
PSO | Robertson and Trucco [95] | Section 10.1 |
Zhang et al. [96] | Section 10.1 | |
Panteleris and Argyros [97] | Section 10.1 | |
Ye et al. [102] | Section 10.2 | |
Panteleris and Argyros [103] | Section 10.2 | |
Oikonomidis et al. [104] | Section 10.2 | |
Padeleris et al. [106] | Section 10.3 |
11 Others
12 Conclusion
13 \thelikesection Appendix A: Pseudo-codes
13.1 \thelikesubsection GA
-
selectParents() (Algorithm 1, line 6)
-
crossover() (Algorithm 1, line 7)
-
mutate() (Algorithm 1, line 10)
13.2 \thelikesubsection DE
-
mutate() (Algorithm 2, line 5)
-
crossover() (Algorithm 2, line 6)
13.3 \thelikesubsection PSO
-
adjustVelocity() (Algorithm 3, line 8)
-
adjustPosition() (Algorithm 3, line 9)
13.4 \thelikesubsection ACO
-
constructSolution() (Algorithm 4, line 4)
-
daemonActions() (Algorithm 4, line 6)
-
updatePheromones() (Algorithm 4, line 8)