• DeepFace Repository A Lightweight Face Recognition and Facial Attribute Analysis (Age, Gender, Emotion and Race) Library for Python.

• Github Repository

DeepFace

docker build -t tensorflow/deepface .

docker run --ipc=host --gpus all -ti --rm \
    -v $(pwd)/notebooks:/opt/app/notebooks \
    -v $(pwd)/weights:/root/.deepface/weights \
    -p 8888:8888 --name deepface \
    tensorflow/deepface

Comparing a few models for face detection and recognition with deepface:

• DeepFace
  • Face Match
  • Compare Image to a set of Images
    • VGG-Face / OpenCV
    • Dlib / dlib
    • Dlib / Mediapipe
    • Dlib / YOLOv8
    • VGG-Face / YOLOv8
    • Facenet512 / YOLOv8
    • Facenet / YOLOv8
    • ArcFace / YOLOv8
    • GhostFaceNet / YOLOv8
  • Face Analysis

from deepface import DeepFace
import glob
import os
import pandas as pd
import matplotlib.pyplot as plt

images = glob.glob('./assets/*.jpg')

plt.figure(figsize=(12,10))

for i in range(25):
    ax = plt.subplot(5,5,i+1)
    plt.title(images[i][9:-4])
    img = plt.imread(images[i])
    plt.imshow(img)
    plt.axis('off')

Face Match

Compare an image to a reference image.

Variable	Description	Options
model_name	Model for face recognition.	VGG-Face, Facenet, Facenet512, OpenFace, DeepFace, DeepID, Dlib, ArcFace, SFace, GhostFaceNet (default is VGG-Face).
detector_backend	Face detector backend.	opencv, retinaface, mtcnn, ssd, dlib, mediapipe, yolov8, centerface, skip (default is opencv).
distance_metric	Metric for measuring similarity.	cosine, euclidean, euclidean_l2 (default is cosine).

result1 = DeepFace.verify(
    img1_path = "assets/Benedict_Wong_01.jpg",
    img2_path = "assets/Benedict_Wong_03.jpg",
    model_name = "VGG-Face",
    detector_backend = "opencv",
    distance_metric = "cosine",
    enforce_detection = True,
    align = True,
    expand_percentage = 0,
    normalization = "base",
    silent = True,
    threshold = 0.75,
) 

print(pretty_print(result1))

Match: True, Distance: 0.2371475924188441, Threshold: 0.75, Time: 0.43, Detections :: x,y,w,h,left_eye,right_eye IMG1: 149,79,167,167,(260, 144),(199, 146) IMG2: 77,85,157,157,None,None None

result2 = DeepFace.verify(
    img1_path = "assets/Benedict_Wong_01.jpg",
    img2_path = "assets/Benedict_Wong_02.jpg",
    model_name = "VGG-Face",
    detector_backend = "opencv",
    distance_metric = "cosine",
    enforce_detection = True,
    align = True,
    expand_percentage = 0,
    normalization = "base",
    silent = True,
    threshold = 0.75,
) 

print(pretty_print(result2))

Match: True, Distance: 0.36789182962805234, Threshold: 0.75, Time: 0.4, Detections :: x,y,w,h,left_eye,right_eye IMG1: 149,79,167,167,(260, 144),(199, 146) IMG2: 60,24,52,52,None,None None

result3 = DeepFace.verify(
    img1_path = "assets/Benedict_Wong_01.jpg",
    img2_path = "assets/John_Bradley_04.jpg",
    model_name = "VGG-Face",
    detector_backend = "opencv",
    distance_metric = "cosine",
    enforce_detection = True,
    align = True,
    expand_percentage = 0,
    normalization = "base",
    silent = True,
    threshold = 0.75,
) 

print(pretty_print(result3))

Match: False, Distance: 0.8878862927720679, Threshold: 0.75, Time: 0.41, Detections :: x,y,w,h,left_eye,right_eye IMG1: 149,79,167,167,(260, 144),(199, 146) IMG2: 25,74,185,185,(151, 149),(83, 150) None

Compare Image to a set of Images

VGG-Face / OpenCV

• vgg_face_weights.h5 - 552.3 MB
• 24-05-09 12:32:58 - Searching assets/Eiza_Gonzalez_03.jpg in 26 length datastore
• 24-05-09 12:32:58 - Find function duration 0.21913480758666992 seconds

result4 = DeepFace.find(
    img_path = "assets/Eiza_Gonzalez_03.jpg",
    db_path = "assets",
    model_name = "VGG-Face",
    distance_metric = "cosine",
    enforce_detection = True,
    detector_backend = "opencv",
    align = True,
    expand_percentage = 0,
    threshold = 0.9,
    normalization = "base",
    silent = False,
)

24-05-09 12:32:58 - Searching assets/Eiza_Gonzalez_03.jpg in 26 length datastore 24-05-09 12:32:58 - find function duration 0.21913480758666992 seconds

#	identity	hash	target_x	target_y	target_w	target_h	source_x	source_y	source_w	source_h	threshold	distance
0	assets/Eiza_Gonzalez_03.jpg	c851bc623b011079548c899eedc96a0b8c1a9264	42	27	71	71	42	27	71	71	0.9	-2.220446e-16
1	assets/Eiza_Gonzalez_05.jpg	4e3c950ef9084dd264e056711afd7a949dfb5f95	49	61	148	148	42	27	71	71	0.9	5.568067e-01
2	assets/John_Bradley_01.jpg	7619cdf25b6b4071aa6ebd9408e001f372cee5bc	69	142	236	236	42	27	71	71	0.9	8.328525e-01
3	assets/John_Bradley_04.jpg	d4c9635f2c37df80ade71f405b296ac98d68e690	25	74	185	185	42	27	71	71	0.9	8.535972e-01
4	assets/Eiza_Gonzalez_02.jpg	220a99ec5380b5d4033dd18f6f54353a49bb341b	462	125	246	246	42	27	71	71	0.9	8.668110e-01
5	assets/Jess_Hong_02.jpg	1233412de754501680ab281fdd0b3c0b7e634279	34	19	81	81	42	27	71	71	0.9	8.805470e-01

image_results = [
    ['./assets/Eiza_Gonzalez_03.jpg',  1.0000000000000002],
    ['./assets/Eiza_Gonzalez_05.jpg',  0.4431933],
    ['./assets/John_Bradley_01.jpg',  0.1671475],
    ['./assets/John_Bradley_04.jpg',  0.14640280000000006],
    ['./assets/Eiza_Gonzalez_02.jpg',  0.133189],
    ['./assets/Jess_Hong_02.jpg',  0.11945300000000003]
]

plt.figure(figsize=(12,8))

for i in range(len(image_results)):
    ax = plt.subplot(2,3,i+1)
    plt.title(
        image_results[i][0][9:-4]+'\n Similarity: '+str(round(image_results[i][1],2))
    )
    img = plt.imread(image_results[i][0])
    plt.imshow(img)
    plt.axis('off')

data_images = [
    'Eiza_Gonzalez_03.jpg',
    'Eiza_Gonzalez_05.jpg',
    'John_Bradley_01.jpg',
    'John_Bradley_04.jpg',
    'Eiza_Gonzalez_02.jpg',
    'Jess_Hong_02.jpg'
]

distance_data = [
     -2.220446e-16,
     5.568067e-01,
     8.328525e-01,
     8.535972e-01,
     8.668110e-01,
     8.805470e-01
]

similarity_data = [
     1.0000000000000002,
     0.4431933,
     0.1671475,
     0.14640280000000006,
     0.133189,
     0.11945300000000003
]

result_4_df = pd.DataFrame({ 
    'Images': data_images, 
    'Distance': distance_data, 
    'Similarity': similarity_data  
}) 

result_4_df

	Images	Distance	Similarity
0	Eiza_Gonzalez_03.jpg	-2.220446e-16	1.000000
1	Eiza_Gonzalez_05.jpg	5.568067e-01	0.443193
2	John_Bradley_01.jpg	8.328525e-01	0.167148
3	John_Bradley_04.jpg	8.535972e-01	0.146403
4	Eiza_Gonzalez_02.jpg	8.668110e-01	0.133189
5	Jess_Hong_02.jpg	8.805470e-01	0.119453

result_4_df.plot(x="Images", y="Similarity", kind="bar")

Dlib / dlib

• dlib_face_recognition_resnet_model_v1.dat.bz2 - 20.4 MB
• 24-05-09 10:29:22 - Searching assets/Eiza_Gonzalez_03.jpg in 26 length datastore
• 24-05-09 10:29:23 - Find function duration 0.13037323951721191 seconds

result5 = DeepFace.find(
    img_path = "assets/Eiza_Gonzalez_03.jpg",
    db_path = "assets",
    model_name = "Dlib",
    distance_metric = "cosine",
    enforce_detection = True,
    detector_backend = "dlib",
    align = True,
    expand_percentage = 0,
    threshold = 0.60,
    normalization = "base",
    silent = False,
)

24-05-09 10:29:22 - Searching assets/Eiza_Gonzalez_03.jpg in 26 length datastore 24-05-09 10:29:23 - find function duration 0.13037323951721191 seconds

#	identity	hash	target_x	target_y	target_w	target_h	source_x	source_y	source_w	source_h	threshold	distance
0	assets/Eiza_Gonzalez_03.jpg	c851bc623b011079548c899eedc96a0b8c1a9264	45	32	62	62	45	32	62	62	0.6	-2.220446e-16
1	assets/Eiza_Gonzalez_01.jpg	dd70311d238d2be8b47aee363e3f228b741c3b50	201	81	155	155	45	32	62	62	0.6	3.507291e-02
2	assets/Eiza_Gonzalez_04.jpg	863a9e1f9f06ff1da45ef39de64f03a24d768164	24	53	63	62	45	32	62	62	0.6	5.434244e-02
3	assets/Eiza_Gonzalez_05.jpg	4e3c950ef9084dd264e056711afd7a949dfb5f95	53	82	129	129	45	32	62	62	0.6	7.985102e-02
4	assets/Eiza_Gonzalez_02.jpg	220a99ec5380b5d4033dd18f6f54353a49bb341b	464	167	223	223	45	32	62	62	0.6	8.234110e-02
5	assets/Alex_Sharp_02.jpg	c95ee8e7698895fe15c561a9cb5adeda43f56ef4	46	46	89	90	45	32	62	62	0.6	1.173593e-01
6	assets/Jess_Hong_01.jpg	46312c1e70040518e2e9ed9e6068cefed5724c08	95	61	52	52	45	32	62	62	0.6	1.268085e-01
7	assets/Jess_Hong_02.jpg	1233412de754501680ab281fdd0b3c0b7e634279	38	30	75	75	45	32	62	62	0.6	1.458864e-01
8	assets/John_Bradley_03.jpg	9686d791c5a87ff6060e721caf66f2532f92300a	44	32	107	108	45	32	62	62	0.6	1.475067e-01
9	assets/John_Bradley_02.jpg	7f0317379f612eb44096207202e3cc91e441a58e	107	66	63	63	45	32	62	62	0.6	1.524923e-01
10	assets/John_Bradley_01.jpg	7619cdf25b6b4071aa6ebd9408e001f372cee5bc	68	142	223	223	45	32	62	62	0.6	1.616775e-01
11	assets/Benedict_Wong_04.jpg	dbb025ab8039915335d84700c6423fadfa91c5b3	11	46	62	62	45	32	62	62	0.6	1.624690e-01
12	assets/Jess_Hong_03.jpg	9b056b12303c62eeee91c49456bdda7f494aa078	199	44	107	108	45	32	62	62	0.6	1.630522e-01
13	assets/Alex_Sharp_01.jpg	7155ec0a507fd98583c179beb164c9fcfa0fa917	211	44	107	108	45	32	62	62	0.6	1.663933e-01
14	assets/Alex_Sharp_03.jpg	48a617ed38787e3dd5e7d7e87e2a56b37214ce38	82	142	267	267	45	32	62	62	0.6	1.686072e-01
15	assets/Jovan_Adepo_01.jpg	47bc06e224a2538aa258c0b5d03f478e5713b8bc	38	30	75	75	45	32	62	62	0.6	1.723205e-01
16	assets/Liam_Cunningham_04.jpg	48cfc1ce0fec1aa61293d5d95478c694fa840686	56	77	165	186	45	32	62	62	0.6	1.757963e-01
17	assets/John_Bradley_04.jpg	d4c9635f2c37df80ade71f405b296ac98d68e690	15	98	186	186	45	32	62	62	0.6	1.834919e-01
18	assets/Liam_Cunningham_02.jpg	5b5aa2bddb00a676c599ab365031336ec1c9e00e	0	26	86	90	45	32	62	62	0.6	1.899648e-01
19	assets/Benedict_Wong_02.jpg	f8aa9b6edaf2e71f3c69cec9548c27519ba1914c	60	26	52	52	45	32	62	62	0.6	1.958830e-01
20	assets/Liam_Cunningham_01.jpg	a157a9143b88b9844ea65f1ccea484a337670853	98	170	321	321	45	32	62	62	0.6	1.977721e-01
21	assets/Benedict_Wong_05.jpg	4f14065c9065e745f63b1ca22c206649922eaee2	40	26	44	44	45	32	62	62	0.6	2.021431e-01
22	assets/Jovan_Adepo_02.jpg	f5428170e3323a333121ceb196675526d8ad02c6	67	82	129	129	45	32	62	62	0.6	2.065918e-01
23	assets/Liam_Cunningham_03.jpg	d349c11690daa1a9e118df5d4ae08d554852a96a	98	81	155	155	45	32	62	62	0.6	2.114577e-01
24	assets/Benedict_Wong_01.jpg	5594e234a03a4a8fb8c8c2a29310cb6497635317	167	110	129	129	45	32	62	62	0.6	2.213369e-01
25	assets/Benedict_Wong_03.jpg	7626233c27c3925bb4d15b301f525ec82e14c868	96	110	129	129	45	32	62	62	0.6	2.251138e-01

image_results = [
    ['./assets/Eiza_Gonzalez_03.jpg',  1.0000000000000002],
    ['./assets/Eiza_Gonzalez_01.jpg',  0.96492709],
    ['./assets/Eiza_Gonzalez_04.jpg',  0.94565756],
    ['./assets/Eiza_Gonzalez_05.jpg',  0.92014898],
    ['./assets/Eiza_Gonzalez_02.jpg',  0.9176588999999999],
    ['./assets/Alex_Sharp_02.jpg',  0.8826407000000001]
]

plt.figure(figsize=(12,8))

for i in range(len(image_results)):
    ax = plt.subplot(2,3,i+1)
    plt.title(
        image_results[i][0][9:-4]+'\n Similarity: '+str(round(image_results[i][1],2))
    )
    img = plt.imread(image_results[i][0])
    plt.imshow(img)
    plt.axis('off')

data_images = [
    'Eiza_Gonzalez_03.jpg',
    'Eiza_Gonzalez_01.jpg',
    'Eiza_Gonzalez_04.jpg',
    'Eiza_Gonzalez_05.jpg',
    'Eiza_Gonzalez_02.jpg',
    'Alex_Sharp_02.jpg',
    'Jess_Hong_01.jpg',
    'Jess_Hong_02.jpg',
    'John_Bradley_03.jpg',
    'John_Bradley_02.jpg',
    'John_Bradley_01.jpg',
    'Benedict_Wong_04.jpg',
    'Jess_Hong_03.jpg',
    'Alex_Sharp_01.jpg',
    'Alex_Sharp_03.jpg',
    'Jovan_Adepo_01.jpg',
    'Liam_Cunningham_04.jpg',
    'John_Bradley_04.jpg',
    'Liam_Cunningham_02.jpg',
    'Benedict_Wong_02.jpg',
    'Liam_Cunningham_01.jpg',
    'Benedict_Wong_05.jpg',
    'Jovan_Adepo_02.jpg',
    'Liam_Cunningham_03.jpg',
    'Benedict_Wong_01.jpg',
    'Benedict_Wong_03.jpg'
]

distance_data = [
     -2.220446e-16,
     3.507291e-02,
     5.434244e-02,
     7.985102e-02,
     8.234110e-02,
     1.173593e-01,
     1.268085e-01,
     1.458864e-01,
     1.475067e-01,
     1.524923e-01,
     1.616775e-01,
     1.624690e-01,
     1.630522e-01,
     1.663933e-01,
     1.686072e-01,
     1.723205e-01,
     1.757963e-01,
     1.834919e-01,
     1.899648e-01,
     1.958830e-01,
     1.977721e-01,
     2.021431e-01,
     2.065918e-01,
     2.114577e-01,
     2.213369e-01,
     2.251138e-01,
]

similarity_data = [
     1.0000000000000002,
     0.96492709,
     0.94565756,
     0.92014898,
     0.9176588999999999,
     0.8826407000000001,
     0.8731915,
     0.8541136,
     0.8524933,
     0.8475077,
     0.8383225,
     0.837531,
     0.8369478,
     0.8336067,
     0.8313927999999999,
     0.8276795,
     0.8242037,
     0.8165081,
     0.8100352,
     0.804117,
     0.8022279,
     0.7978569,
     0.7934082,
     0.7885423,
     0.7786630999999999,
     0.7748862,
]

result_5_df = pd.DataFrame({ 
    'Images': data_images, 
    'Distance': distance_data, 
    'Similarity': similarity_data  
}) 

result_5_df

	Images	Distance	Similarity
0	Eiza_Gonzalez_03.jpg	-2.220446e-16	1.000000
1	Eiza_Gonzalez_01.jpg	3.507291e-02	0.964927
2	Eiza_Gonzalez_04.jpg	5.434244e-02	0.945658
3	Eiza_Gonzalez_05.jpg	7.985102e-02	0.920149
4	Eiza_Gonzalez_02.jpg	8.234110e-02	0.917659
5	Alex_Sharp_02.jpg	1.173593e-01	0.882641
6	Jess_Hong_01.jpg	1.268085e-01	0.873192
7	Jess_Hong_02.jpg	1.458864e-01	0.854114
8	John_Bradley_03.jpg	1.475067e-01	0.852493
9	John_Bradley_02.jpg	1.524923e-01	0.847508
10	John_Bradley_01.jpg	1.616775e-01	0.838322
11	Benedict_Wong_04.jpg	1.624690e-01	0.837531
12	Jess_Hong_03.jpg	1.630522e-01	0.836948
13	Alex_Sharp_01.jpg	1.663933e-01	0.833607
14	Alex_Sharp_03.jpg	1.686072e-01	0.831393
15	Jovan_Adepo_01.jpg	1.723205e-01	0.827680
16	Liam_Cunningham_04.jpg	1.757963e-01	0.824204
17	John_Bradley_04.jpg	1.834919e-01	0.816508
18	Liam_Cunningham_02.jpg	1.899648e-01	0.810035
19	Benedict_Wong_02.jpg	1.958830e-01	0.804117
20	Liam_Cunningham_01.jpg	1.977721e-01	0.802228
21	Benedict_Wong_05.jpg	2.021431e-01	0.797857
22	Jovan_Adepo_02.jpg	2.065918e-01	0.793408
23	Liam_Cunningham_03.jpg	2.114577e-01	0.788542
24	Benedict_Wong_01.jpg	2.213369e-01	0.778663
25	Benedict_Wong_03.jpg	2.251138e-01	0.774886

def pretty_print(result):
 print(f"Match: {result['verified']},\nDistance: {result['distance']},\nThreshold: {result['threshold']},\nTime: {result['time']},\nDetections :: x,y,w,h,left_eye,right_eye\n    IMG1: {result['facial_areas']['img1']['x']},{result['facial_areas']['img1']['y']},{result['facial_areas']['img1']['w']},{result['facial_areas']['img1']['h']},{result['facial_areas']['img1']['left_eye']},{result['facial_areas']['img1']['right_eye']}\n    IMG2: {result['facial_areas']['img2']['x']},{result['facial_areas']['img2']['y']},{result['facial_areas']['img2']['w']},{result['facial_areas']['img2']['h']},{result['facial_areas']['img2']['left_eye']},{result['facial_areas']['img2']['right_eye']}")

Dlib / Mediapipe

• dlib_face_recognition_resnet_model_v1.dat.bz2 - 20.4 MB
• shape_predictor_5_face_landmarks.dat.bz2 - 5.4 MB
• 24-05-10 07:38:39 - Searching assets/Eiza_Gonzalez_03.jpg in 27 length datastore
• 24-05-10 07:38:39 - Find function duration 0.3898661136627197 seconds

result6 = DeepFace.find(
    img_path = "assets/Eiza_Gonzalez_03.jpg",
    db_path = "assets",
    model_name = "Dlib",
    distance_metric = "cosine",
    enforce_detection = True,
    detector_backend = "mediapipe",
    align = True,
    expand_percentage = 0,
    threshold = 0.60,
    normalization = "base",
    silent = False,
)

24-05-10 07:38:39 - Searching assets/Eiza_Gonzalez_03.jpg in 27 length datastore 24-05-10 07:38:39 - find function duration 0.3898661136627197 seconds

INFO: Created TensorFlow Lite XNNPACK delegate for CPU.

#	identity	hash	target_x	target_y	target_w	target_h	source_x	source_y	source_w	source_h	threshold	distance
0	assets/Eiza_Gonzalez_03.jpg	c851bc623b011079548c899eedc96a0b8c1a9264	37	36	65	65	37	36	65	65	0.6	0.000000
1	assets/Eiza_Gonzalez_02.jpg	220a99ec5380b5d4033dd18f6f54353a49bb341b	509	181	197	197	37	36	65	65	0.6	0.058698
2	assets/Eiza_Gonzalez_04.jpg	863a9e1f9f06ff1da45ef39de64f03a24d768164	16	41	80	80	37	36	65	65	0.6	0.069220
3	assets/Eiza_Gonzalez_05.jpg	4e3c950ef9084dd264e056711afd7a949dfb5f95	49	84	133	133	37	36	65	65	0.6	0.070974
4	assets/Jess_Hong_01.jpg	46312c1e70040518e2e9ed9e6068cefed5724c08	113	71	48	48	37	36	65	65	0.6	0.116602
5	assets/Jess_Hong_02.jpg	1233412de754501680ab281fdd0b3c0b7e634279	39	28	75	75	37	36	65	65	0.6	0.122306
6	assets/John_Bradley_03.jpg	9686d791c5a87ff6060e721caf66f2532f92300a	32	36	104	104	37	36	65	65	0.6	0.127752
7	assets/Alex_Sharp_02.jpg	c95ee8e7698895fe15c561a9cb5adeda43f56ef4	38	47	101	101	37	36	65	65	0.6	0.145324
8	assets/Alex_Sharp_01.jpg	7155ec0a507fd98583c179beb164c9fcfa0fa917	202	57	107	107	37	36	65	65	0.6	0.148071
9	assets/John_Bradley_02.jpg	7f0317379f612eb44096207202e3cc91e441a58e	120	77	53	53	37	36	65	65	0.6	0.154632
10	assets/Liam_Cunningham_02.jpg	5b5aa2bddb00a676c599ab365031336ec1c9e00e	12	29	84	84	37	36	65	65	0.6	0.163341
11	assets/John_Bradley_01.jpg	7619cdf25b6b4071aa6ebd9408e001f372cee5bc	96	184	208	208	37	36	65	65	0.6	0.164641
12	assets/Alex_Sharp_03.jpg	48a617ed38787e3dd5e7d7e87e2a56b37214ce38	106	156	267	267	37	36	65	65	0.6	0.164873
13	assets/Eiza_Gonzalez_02.jpg	220a99ec5380b5d4033dd18f6f54353a49bb341b	55	271	201	201	37	36	65	65	0.6	0.168053
14	assets/Liam_Cunningham_04.jpg	48cfc1ce0fec1aa61293d5d95478c694fa840686	44	99	151	151	37	36	65	65	0.6	0.171255
15	assets/Benedict_Wong_04.jpg	dbb025ab8039915335d84700c6423fadfa91c5b3	13	46	56	56	37	36	65	65	0.6	0.178435
16	assets/Benedict_Wong_02.jpg	f8aa9b6edaf2e71f3c69cec9548c27519ba1914c	64	34	43	43	37	36	65	65	0.6	0.186274
17	assets/Benedict_Wong_05.jpg	4f14065c9065e745f63b1ca22c206649922eaee2	41	29	40	40	37	36	65	65	0.6	0.191297
18	assets/John_Bradley_04.jpg	d4c9635f2c37df80ade71f405b296ac98d68e690	27	111	168	168	37	36	65	65	0.6	0.197620
19	assets/Liam_Cunningham_01.jpg	a157a9143b88b9844ea65f1ccea484a337670853	114	215	262	262	37	36	65	65	0.6	0.202479
20	assets/Jovan_Adepo_02.jpg	f5428170e3323a333121ceb196675526d8ad02c6	59	96	125	125	37	36	65	65	0.6	0.204949
21	assets/Benedict_Wong_01.jpg	5594e234a03a4a8fb8c8c2a29310cb6497635317	150	107	160	160	37	36	65	65	0.6	0.223669
22	assets/Jovan_Adepo_01.jpg	47bc06e224a2538aa258c0b5d03f478e5713b8bc	42	36	71	71	37	36	65	65	0.6	0.246249

image_results = [
    ['./assets/Eiza_Gonzalez_03.jpg',  1.000000000000000],
    ['./assets/Eiza_Gonzalez_02.jpg',  0.941302],
    ['./assets/Eiza_Gonzalez_04.jpg',  0.9307799999999999],
    ['./assets/Eiza_Gonzalez_05.jpg',  0.929026],
    ['./assets/Jess_Hong_01.jpg',  0.883398],
    ['./assets/Jess_Hong_02.jpg',  0.877694]
]

plt.figure(figsize=(12,8))

for i in range(len(image_results)):
    ax = plt.subplot(2,3,i+1)
    plt.title(
        image_results[i][0][9:-4]+'\n Similarity: '+str(round(image_results[i][1],2))
    )
    img = plt.imread(image_results[i][0])
    plt.imshow(img)
    plt.axis('off')

data_images = [
    'Eiza_Gonzalez_03.jpg',
    'Eiza_Gonzalez_02.jpg',
    'Eiza_Gonzalez_04.jpg',
    'Eiza_Gonzalez_05.jpg',
    'Jess_Hong_01.jpg',
    'Jess_Hong_02.jpg',
    'John_Bradley_03.jpg',
    'Alex_Sharp_02.jpg',
    'Alex_Sharp_01.jpg',
    'John_Bradley_02.jpg',
    'Liam_Cunningham_02.jpg',
    'John_Bradley_01.jpg',
    'Alex_Sharp_03.jpg',
    'Eiza_Gonzalez_02.jpg',
    'Liam_Cunningham_04.jpg',
    'Benedict_Wong_04.jpg',
    'Benedict_Wong_02.jpg',
    'Benedict_Wong_05.jpg',
    'John_Bradley_04.jpg',
    'Liam_Cunningham_01.jpg',
    'Jovan_Adepo_02.jpg',
    'Benedict_Wong_01.jpg',
    'Jovan_Adepo_01.jpg'
]

distance_data = [
     0.000000,
     0.058698,
     0.069220,
     0.070974,
     0.116602,
     0.122306,
     0.127752,
     0.145324,
     0.148071,
     0.154632,
     0.163341,
     0.164641,
     0.164873,
     0.168053,
     0.171255,
     0.178435,
     0.186274,
     0.191297,
     0.197620,
     0.202479,
     0.204949,
     0.223669,
     0.246249
]

similarity_data = [
     1.0,
     0.941302,
     0.9307799999999999,
     0.929026,
     0.883398,
     0.877694,
     0.872248,
     0.854676,
     0.8519289999999999,
     0.845368,
     0.836659,
     0.835359,
     0.835127,
     0.831947,
     0.8287450000000001,
     0.821565,
     0.813726,
     0.808703,
     0.80238,
     0.797521,
     0.795051,
     0.776331,
     0.7537510000000001
]

result_6_df = pd.DataFrame({ 
    'Images': data_images, 
    'Distance': distance_data, 
    'Similarity': similarity_data  
}) 

result_6_df

	Images	Distance	Similarity
0	Eiza_Gonzalez_03.jpg	0.000000	1.000000
1	Eiza_Gonzalez_02.jpg	0.058698	0.941302
2	Eiza_Gonzalez_04.jpg	0.069220	0.930780
3	Eiza_Gonzalez_05.jpg	0.070974	0.929026
4	Jess_Hong_01.jpg	0.116602	0.883398
5	Jess_Hong_02.jpg	0.122306	0.877694
6	John_Bradley_03.jpg	0.127752	0.872248
7	Alex_Sharp_02.jpg	0.145324	0.854676
8	Alex_Sharp_01.jpg	0.148071	0.851929
9	John_Bradley_02.jpg	0.154632	0.845368
10	Liam_Cunningham_02.jpg	0.163341	0.836659
11	John_Bradley_01.jpg	0.164641	0.835359
12	Alex_Sharp_03.jpg	0.164873	0.835127
13	Eiza_Gonzalez_02.jpg	0.168053	0.831947
14	Liam_Cunningham_04.jpg	0.171255	0.828745
15	Benedict_Wong_04.jpg	0.178435	0.821565
16	Benedict_Wong_02.jpg	0.186274	0.813726
17	Benedict_Wong_05.jpg	0.191297	0.808703
18	John_Bradley_04.jpg	0.197620	0.802380
19	Liam_Cunningham_01.jpg	0.202479	0.797521
20	Jovan_Adepo_02.jpg	0.204949	0.795051
21	Benedict_Wong_01.jpg	0.223669	0.776331
22	Jovan_Adepo_01.jpg	0.246249	0.753751

result_6_df.plot(x="Images", y="Similarity", kind="bar")

Dlib / YOLOv8

• dlib_face_recognition_resnet_model_v1.dat.bz2 - 20.4 MB
• yolov8n-face.pt - 6.1 MB
• 24-05-10 07:37:37 - Searching assets/Eiza_Gonzalez_03.jpg in 27 length datastore
• 24-05-10 07:37:37 - Find function duration 0.12443137168884277 seconds

result7 = DeepFace.find(
    img_path = "assets/Eiza_Gonzalez_03.jpg",
    db_path = "assets",
    model_name = "Dlib",
    distance_metric = "cosine",
    enforce_detection = True,
    detector_backend = "yolov8",
    align = True,
    expand_percentage = 0,
    threshold = 0.60,
    normalization = "base",
    silent = False,
)

24-05-10 07:37:37 - Searching assets/Eiza_Gonzalez_03.jpg in 27 length datastore 24-05-10 07:37:37 - find function duration 0.12443137168884277 seconds

#	identity	hash	target_x	target_y	target_w	target_h	source_x	source_y	source_w	source_h	threshold	distance
0	assets/Eiza_Gonzalez_03.jpg	c851bc623b011079548c899eedc96a0b8c1a9264	39	19	62	84	39	19	62	84	0.6	0.000000
1	assets/Eiza_Gonzalez_01.jpg	dd70311d238d2be8b47aee363e3f228b741c3b50	203	65	139	183	39	19	62	84	0.6	0.045388
2	assets/Eiza_Gonzalez_05.jpg	4e3c950ef9084dd264e056711afd7a949dfb5f95	57	53	119	163	39	19	62	84	0.6	0.046588
3	assets/Jess_Hong_02.jpg	1233412de754501680ab281fdd0b3c0b7e634279	43	14	67	89	39	19	62	84	0.6	0.055280
4	assets/Eiza_Gonzalez_04.jpg	863a9e1f9f06ff1da45ef39de64f03a24d768164	19	23	79	97	39	19	62	84	0.6	0.063742
5	assets/Jess_Hong_01.jpg	46312c1e70040518e2e9ed9e6068cefed5724c08	110	54	48	69	39	19	62	84	0.6	0.066402
6	assets/Alex_Sharp_02.jpg	c95ee8e7698895fe15c561a9cb5adeda43f56ef4	39	16	92	135	39	19	62	84	0.6	0.084785
7	assets/Liam_Cunningham_04.jpg	48cfc1ce0fec1aa61293d5d95478c694fa840686	58	52	140	207	39	19	62	84	0.6	0.086134
8	assets/Jess_Hong_03.jpg	9b056b12303c62eeee91c49456bdda7f494aa078	206	24	101	135	39	19	62	84	0.6	0.086582
9	assets/Eiza_Gonzalez_02.jpg	220a99ec5380b5d4033dd18f6f54353a49bb341b	497	132	192	252	39	19	62	84	0.6	0.086979
10	assets/Benedict_Wong_04.jpg	dbb025ab8039915335d84700c6423fadfa91c5b3	19	31	52	74	39	19	62	84	0.6	0.088260
11	assets/John_Bradley_03.jpg	9686d791c5a87ff6060e721caf66f2532f92300a	36	20	95	134	39	19	62	84	0.6	0.091857
12	assets/Alex_Sharp_01.jpg	7155ec0a507fd98583c179beb164c9fcfa0fa917	207	42	93	123	39	19	62	84	0.6	0.104845
13	assets/John_Bradley_04.jpg	d4c9635f2c37df80ade71f405b296ac98d68e690	31	60	166	214	39	19	62	84	0.6	0.104869
14	assets/Eiza_Gonzalez_02.jpg	220a99ec5380b5d4033dd18f6f54353a49bb341b	69	218	132	157	39	19	62	84	0.6	0.104911
15	assets/John_Bradley_01.jpg	7619cdf25b6b4071aa6ebd9408e001f372cee5bc	99	114	198	282	39	19	62	84	0.6	0.108680
16	assets/Liam_Cunningham_02.jpg	5b5aa2bddb00a676c599ab365031336ec1c9e00e	11	10	80	104	39	19	62	84	0.6	0.111524
17	assets/John_Bradley_02.jpg	7f0317379f612eb44096207202e3cc91e441a58e	119	62	55	75	39	19	62	84	0.6	0.113892
18	assets/Benedict_Wong_05.jpg	4f14065c9065e745f63b1ca22c206649922eaee2	45	18	36	49	39	19	62	84	0.6	0.125120
19	assets/Alex_Sharp_03.jpg	48a617ed38787e3dd5e7d7e87e2a56b37214ce38	110	111	248	345	39	19	62	84	0.6	0.126874
20	assets/Benedict_Wong_02.jpg	f8aa9b6edaf2e71f3c69cec9548c27519ba1914c	64	23	44	54	39	19	62	84	0.6	0.127946
21	assets/Benedict_Wong_01.jpg	5594e234a03a4a8fb8c8c2a29310cb6497635317	166	74	138	185	39	19	62	84	0.6	0.134841
22	assets/Benedict_Wong_03.jpg	7626233c27c3925bb4d15b301f525ec82e14c868	89	84	130	173	39	19	62	84	0.6	0.145167
23	assets/Jovan_Adepo_01.jpg	47bc06e224a2538aa258c0b5d03f478e5713b8bc	46	16	64	92	39	19	62	84	0.6	0.146045
24	assets/Liam_Cunningham_01.jpg	a157a9143b88b9844ea65f1ccea484a337670853	137	163	231	319	39	19	62	84	0.6	0.151728
25	assets/Jovan_Adepo_02.jpg	f5428170e3323a333121ceb196675526d8ad02c6	63	59	117	169	39	19	62	84	0.6	0.153202
26	assets/Liam_Cunningham_03.jpg	d349c11690daa1a9e118df5d4ae08d554852a96a	118	58	130	184	39	19	62	84	0.6	0.157119

image_results = [
    ['./assets/Eiza_Gonzalez_03.jpg',  1.000000000000000],
    ['./assets/Eiza_Gonzalez_01.jpg',  0.954612],
    ['./assets/Eiza_Gonzalez_05.jpg',  0.953412],
    ['./assets/Jess_Hong_02.jpg',  0.94472],
    ['./assets/Eiza_Gonzalez_04.jpg',  0.936258],
    ['./assets/Jess_Hong_01.jpg',  0.933598]
]

plt.figure(figsize=(12,8))

for i in range(len(image_results)):
    ax = plt.subplot(2,3,i+1)
    plt.title(
        image_results[i][0][9:-4]+'\n Similarity: '+str(round(image_results[i][1],2))
    )
    img = plt.imread(image_results[i][0])
    plt.imshow(img)
    plt.axis('off')

data_images = [
    'Eiza_Gonzalez_03.jpg',
    'Eiza_Gonzalez_01.jpg',
    'Eiza_Gonzalez_05.jpg',
    'Jess_Hong_02.jpg',
    'Eiza_Gonzalez_04.jpg',
    'Jess_Hong_01.jpg',
    'Alex_Sharp_02.jpg',
    'Liam_Cunningham_04.jpg',
    'Jess_Hong_03.jpg',
    'Eiza_Gonzalez_02.jpg',
    'Benedict_Wong_04.jpg',
    'John_Bradley_03.jpg',
    'Alex_Sharp_01.jpg',
    'John_Bradley_04.jpg',
    'Eiza_Gonzalez_02.jpg',
    'John_Bradley_01.jpg',
    'Liam_Cunningham_02.jpg',
    'John_Bradley_02.jpg',
    'Benedict_Wong_05.jpg',
    'Alex_Sharp_03.jpg',
    'Benedict_Wong_02.jpg',
    'Benedict_Wong_01.jpg',
    'Benedict_Wong_03.jpg',
    'Jovan_Adepo_01.jpg',
    'Liam_Cunningham_01.jpg',
    'Jovan_Adepo_02.jpg',
    'Liam_Cunningham_03.jpg'
]

distance_data = [
     0.000000,
     0.045388,
     0.046588,
     0.055280,
     0.063742,
     0.066402,
     0.084785,
     0.086134,
     0.086582,
     0.086979,
     0.088260,
     0.091857,
     0.104845,
     0.104869,
     0.104911,
     0.108680,
     0.111524,
     0.113892,
     0.125120,
     0.126874,
     0.127946,
     0.134841,
     0.145167,
     0.146045,
     0.151728,
     0.153202,
     0.157119
]

similarity_data = [
     1,
     0.954612,
     0.953412,
     0.94472,
     0.936258,
     0.933598,
     0.915215,
     0.913866,
     0.913418,
     0.913021,
     0.91174,
     0.908143,
     0.895155,
     0.895131,
     0.895089,
     0.89132,
     0.888476,
     0.886108,
     0.87488,
     0.8731260000000001,
     0.872054,
     0.865159,
     0.854833,
     0.853955,
     0.848272,
     0.8467979999999999,
     0.842881
]

result_7_df = pd.DataFrame({ 
    'Images': data_images, 
    'Distance': distance_data, 
    'Similarity': similarity_data  
}) 

result_7_df

	Images	Distance	Similarity
0	Eiza_Gonzalez_03.jpg	0.000000	1.000000
1	Eiza_Gonzalez_01.jpg	0.045388	0.954612
2	Eiza_Gonzalez_05.jpg	0.046588	0.953412
3	Jess_Hong_02.jpg	0.055280	0.944720
4	Eiza_Gonzalez_04.jpg	0.063742	0.936258
5	Jess_Hong_01.jpg	0.066402	0.933598
6	Alex_Sharp_02.jpg	0.084785	0.915215
7	Liam_Cunningham_04.jpg	0.086134	0.913866
8	Jess_Hong_03.jpg	0.086582	0.913418
9	Eiza_Gonzalez_02.jpg	0.086979	0.913021
10	Benedict_Wong_04.jpg	0.088260	0.911740
11	John_Bradley_03.jpg	0.091857	0.908143
12	Alex_Sharp_01.jpg	0.104845	0.895155
13	John_Bradley_04.jpg	0.104869	0.895131
14	Eiza_Gonzalez_02.jpg	0.104911	0.895089
15	John_Bradley_01.jpg	0.108680	0.891320
16	Liam_Cunningham_02.jpg	0.111524	0.888476
17	John_Bradley_02.jpg	0.113892	0.886108
18	Benedict_Wong_05.jpg	0.125120	0.874880
19	Alex_Sharp_03.jpg	0.126874	0.873126
20	Benedict_Wong_02.jpg	0.127946	0.872054
21	Benedict_Wong_01.jpg	0.134841	0.865159
22	Benedict_Wong_03.jpg	0.145167	0.854833
23	Jovan_Adepo_01.jpg	0.146045	0.853955
24	Liam_Cunningham_01.jpg	0.151728	0.848272
25	Jovan_Adepo_02.jpg	0.153202	0.846798
26	Liam_Cunningham_03.jpg	0.157119	0.842881

result_7_df.plot(x="Images", y="Similarity", kind="bar")

VGG-Face / YOLOv8

• vgg_face_weights.h5 - 552.3 MB
• yolov8n-face.pt - 6.1 MB
• 24-05-09 12:31:57 - Searching assets/Eiza_Gonzalez_03.jpg in 27 length datastore
• 24-05-09 12:31:57 - find function duration 0.21353983879089355 seconds

result8 = DeepFace.find(
    img_path = "assets/Eiza_Gonzalez_03.jpg",
    db_path = "assets",
    model_name = "VGG-Face",
    distance_metric = "cosine",
    enforce_detection = True,
    detector_backend = "yolov8",
    align = True,
    expand_percentage = 0,
    threshold = 0.90,
    normalization = "base",
    silent = False,
)

24-05-09 12:31:57 - Searching assets/Eiza_Gonzalez_03.jpg in 27 length datastore 24-05-09 12:31:57 - find function duration 0.21353983879089355 seconds

#	identity	hash	target_x	target_y	target_w	target_h	source_x	source_y	source_w	source_h	threshold	distance
0	assets/Eiza_Gonzalez_03.jpg	c851bc623b011079548c899eedc96a0b8c1a9264	39	19	62	84	39	19	62	84	0.9	-2.220446e-16
1	assets/Eiza_Gonzalez_01.jpg	dd70311d238d2be8b47aee363e3f228b741c3b50	203	65	139	183	39	19	62	84	0.9	4.798819e-01
2	assets/Eiza_Gonzalez_05.jpg	4e3c950ef9084dd264e056711afd7a949dfb5f95	57	53	119	163	39	19	62	84	0.9	5.464998e-01
3	assets/Eiza_Gonzalez_04.jpg	863a9e1f9f06ff1da45ef39de64f03a24d768164	19	23	79	97	39	19	62	84	0.9	5.515515e-01
4	assets/Eiza_Gonzalez_02.jpg	220a99ec5380b5d4033dd18f6f54353a49bb341b	497	132	192	252	39	19	62	84	0.9	7.510090e-01
5	assets/John_Bradley_01.jpg	7619cdf25b6b4071aa6ebd9408e001f372cee5bc	99	114	198	282	39	19	62	84	0.9	8.351889e-01
6	assets/Jess_Hong_02.jpg	1233412de754501680ab281fdd0b3c0b7e634279	43	14	67	89	39	19	62	84	0.9	8.763548e-01
7	assets/Jess_Hong_03.jpg	9b056b12303c62eeee91c49456bdda7f494aa078	206	24	101	135	39	19	62	84	0.9	8.841409e-01
8	assets/John_Bradley_03.jpg	9686d791c5a87ff6060e721caf66f2532f92300a	36	20	95	134	39	19	62	84	0.9	8.854007e-01
9	assets/John_Bradley_04.jpg	d4c9635f2c37df80ade71f405b296ac98d68e690	31	60	166	214	39	19	62	84	0.9	8.909801e-01
10	assets/Benedict_Wong_05.jpg	4f14065c9065e745f63b1ca22c206649922eaee2	45	18	36	49	39	19	62	84	0.9	8.979046e-01

image_results = [
    ['./assets/Eiza_Gonzalez_03.jpg',  1.0000000000000002],
    ['./assets/Eiza_Gonzalez_01.jpg',  0.5201181],
    ['./assets/Eiza_Gonzalez_05.jpg',  0.4535002],
    ['./assets/Eiza_Gonzalez_04.jpg',  0.4484485],
    ['./assets/Eiza_Gonzalez_02.jpg',  0.24899099999999996],
    ['./assets/John_Bradley_01.jpg',  0.1648111]
]

plt.figure(figsize=(12,8))

for i in range(len(image_results)):
    ax = plt.subplot(2,3,i+1)
    plt.title(
        image_results[i][0][9:-4]+'\n Similarity: '+str(round(image_results[i][1],2))
    )
    img = plt.imread(image_results[i][0])
    plt.imshow(img)
    plt.axis('off')

data_images = [
    'Eiza_Gonzalez_03.jpg',
    'Eiza_Gonzalez_01.jpg',
    'Eiza_Gonzalez_05.jpg',
    'Eiza_Gonzalez_04.jpg',
    'Eiza_Gonzalez_02.jpg',
    'John_Bradley_01.jpg',
    'Jess_Hong_02.jpg',
    'Jess_Hong_03.jpg',
    'John_Bradley_03.jpg',
    'John_Bradley_04.jpg',
    'Benedict_Wong_05.jpg'
]

distance_data = [
     -2.220446e-16,
     4.798819e-01,
     5.464998e-01,
     5.515515e-01,
     7.510090e-01,
     8.351889e-01,
     8.763548e-01,
     8.841409e-01,
     8.854007e-01,
     8.909801e-01,
     8.979046e-01
]

similarity_data = [
     1.0000000000000002,
     0.5201181,
     0.4535002,
     0.4484485,
     0.24899099999999996,
     0.1648111,
     0.12364520000000001,
     0.11585909999999999,
     0.11459929999999996,
     0.10901989999999995,
     0.10209539999999995
]

result_8_df = pd.DataFrame({ 
    'Images': data_images, 
    'Distance': distance_data, 
    'Similarity': similarity_data  
}) Facenet512 / YOLOv8

result_8_df

	Images	Distance	Similarity
0	Eiza_Gonzalez_03.jpg	-2.220446e-16	1.000000
1	Eiza_Gonzalez_01.jpg	4.798819e-01	0.520118
2	Eiza_Gonzalez_05.jpg	5.464998e-01	0.453500
3	Eiza_Gonzalez_04.jpg	5.515515e-01	0.448449
4	Eiza_Gonzalez_02.jpg	7.510090e-01	0.248991
5	John_Bradley_01.jpg	8.351889e-01	0.164811
6	Jess_Hong_02.jpg	8.763548e-01	0.123645
7	Jess_Hong_03.jpg	8.841409e-01	0.115859
8	John_Bradley_03.jpg	8.854007e-01	0.114599
9	John_Bradley_04.jpg	8.909801e-01	0.109020
10	Benedict_Wong_05.jpg	8.979046e-01	0.102095

result_8_df.plot(x="Images", y="Similarity", kind="bar")

Facenet512 / YOLOv8

• facenet512_weights.h5 - 90.6 MB
• yolov8n-face.pt - 6.1 MB
• 24-05-10 06:14:06 - Searching assets/Eiza_Gonzalez_03.jpg in 27 length datastore
• 24-05-10 06:14:07 - find function duration 0.1690361499786377 seconds

result9 = DeepFace.find(
    img_path = "assets/Eiza_Gonzalez_03.jpg",
    db_path = "assets",
    model_name = "Facenet512",
    distance_metric = "cosine",
    enforce_detection = True,
    detector_backend = "yolov8",
    align = True,
    expand_percentage = 0,
    threshold = 0.90,
    normalization = "base",
    silent = False,
)

24-05-10 06:14:06 - Searching assets/Eiza_Gonzalez_03.jpg in 27 length datastore 24-05-10 06:14:07 - find function duration 0.1690361499786377 seconds

#	identity	hash	target_x	target_y	target_w	target_h	source_x	source_y	source_w	source_h	threshold	distance
0	assets/Eiza_Gonzalez_03.jpg	c851bc623b011079548c899eedc96a0b8c1a9264	39	19	62	84	39	19	62	84	0.9	4.440892e-16
1	assets/Eiza_Gonzalez_01.jpg	dd70311d238d2be8b47aee363e3f228b741c3b50	203	65	139	183	39	19	62	84	0.9	2.783114e-01
2	assets/Eiza_Gonzalez_04.jpg	863a9e1f9f06ff1da45ef39de64f03a24d768164	19	23	79	97	39	19	62	84	0.9	3.291848e-01
3	assets/Eiza_Gonzalez_05.jpg	4e3c950ef9084dd264e056711afd7a949dfb5f95	57	53	119	163	39	19	62	84	0.9	3.929502e-01
4	assets/Eiza_Gonzalez_02.jpg	220a99ec5380b5d4033dd18f6f54353a49bb341b	497	132	192	252	39	19	62	84	0.9	4.768916e-01
5	assets/Jovan_Adepo_01.jpg	47bc06e224a2538aa258c0b5d03f478e5713b8bc	46	16	64	92	39	19	62	84	0.9	7.567952e-01
6	assets/Jess_Hong_02.jpg	1233412de754501680ab281fdd0b3c0b7e634279	43	14	67	89	39	19	62	84	0.9	8.551544e-01
7	assets/Benedict_Wong_04.jpg	dbb025ab8039915335d84700c6423fadfa91c5b3	19	31	52	74	39	19	62	84	0.9	8.763277e-01
8	assets/Jovan_Adepo_02.jpg	f5428170e3323a333121ceb196675526d8ad02c6	63	59	117	169	39	19	62	84	0.9	8.765136e-01
9	assets/Alex_Sharp_03.jpg	48a617ed38787e3dd5e7d7e87e2a56b37214ce38	110	111	248	345	39	19	62	84	0.9	8.860148e-01
10	assets/John_Bradley_03.jpg	9686d791c5a87ff6060e721caf66f2532f92300a	36	20	95	134	39	19	62	84	0.9	8.862074e-01
11	assets/Liam_Cunningham_03.jpg	d349c11690daa1a9e118df5d4ae08d554852a96a	118	58	130	184	39	19	62	84	0.9	8.896595e-01
12	assets/John_Bradley_02.jpg	7f0317379f612eb44096207202e3cc91e441a58e	119	62	55	75	39	19	62	84	0.9	8.998877e-01

image_results = [
    ['./assets/Eiza_Gonzalez_03.jpg',  0.9999999999999996],
    ['./assets/Eiza_Gonzalez_01.jpg',  0.7216886],
    ['./assets/Eiza_Gonzalez_04.jpg',  0.6708152000000001],
    ['./assets/Eiza_Gonzalez_05.jpg',  0.6070498],
    ['./assets/Eiza_Gonzalez_02.jpg',  0.5231083999999999],
    ['./assets/Jovan_Adepo_01.jpg',  0.2432048]
]

plt.figure(figsize=(12,8))

for i in range(len(image_results)):
    ax = plt.subplot(2,3,i+1)
    plt.title(
        image_results[i][0][9:-4]+'\n Similarity: '+str(round(image_results[i][1],2))
    )
    img = plt.imread(image_results[i][0])
    plt.imshow(img)
    plt.axis('off')

data_images = [
    'Eiza_Gonzalez_03.jpg',
    'Eiza_Gonzalez_01.jpg',
    'Eiza_Gonzalez_04.jpg',
    'Eiza_Gonzalez_05.jpg',
    'Eiza_Gonzalez_02.jpg',
    'Jovan_Adepo_01.jpg',
    'Jess_Hong_02.jpg',
    'Benedict_Wong_04.jpg',
    'Jovan_Adepo_02.jpg',
    'Alex_Sharp_03.jpg',
    'John_Bradley_03.jpg',
    'Liam_Cunningham_03.jpg',
    'John_Bradley_02.jpg'
]

distance_data = [
     4.440892e-16,
     2.783114e-01,
     3.291848e-01,
     3.929502e-01,
     4.768916e-01,
     7.567952e-01,
     8.551544e-01,
     8.763277e-01,
     8.765136e-01,
     8.860148e-01,
     8.862074e-01,
     8.896595e-01,
     8.998877e-01,
]

similarity_data = [
     0.9999999999999996,
     0.7216886,
     0.6708152000000001,
     0.6070498,
     0.5231083999999999,
     0.2432048,
     0.14484560000000002,
     0.12367229999999996,
     0.1234864,
     0.11398520000000001,
     0.11379260000000002,
     0.11034049999999995,
     0.10011230000000004
]

result_9_df = pd.DataFrame({ 
    'Images': data_images, 
    'Distance': distance_data, 
    'Similarity': similarity_data  
}) 

result_9_df

	Images	Distance	Similarity
0	Eiza_Gonzalez_03.jpg	4.440892e-16	1.000000
1	Eiza_Gonzalez_01.jpg	2.783114e-01	0.721689
2	Eiza_Gonzalez_04.jpg	3.291848e-01	0.670815
3	Eiza_Gonzalez_05.jpg	3.929502e-01	0.607050
4	Eiza_Gonzalez_02.jpg	4.768916e-01	0.523108
5	Jovan_Adepo_01.jpg	7.567952e-01	0.243205
6	Jess_Hong_02.jpg	8.551544e-01	0.144846
7	Benedict_Wong_04.jpg	8.763277e-01	0.123672
8	Jovan_Adepo_02.jpg	8.765136e-01	0.123486
9	Alex_Sharp_03.jpg	8.860148e-01	0.113985
10	John_Bradley_03.jpg	8.862074e-01	0.113793
11	Liam_Cunningham_03.jpg	8.896595e-01	0.110340
12	John_Bradley_02.jpg	8.998877e-01	0.100112

result_9_df.plot(x="Images", y="Similarity", kind="bar")

Facenet / YOLOv8

• facenet_weights.h5 - 87.9 MB
• yolov8n-face.pt - 6.1 MB
• 24-05-10 06:31:52 - Searching assets/Eiza_Gonzalez_03.jpg in 27 length datastore
• 24-05-10 06:31:53 - find function duration 0.1595611572265625 seconds

result10 = DeepFace.find(
    img_path = "assets/Eiza_Gonzalez_03.jpg",
    db_path = "assets",
    model_name = "Facenet",
    distance_metric = "cosine",
    enforce_detection = True,
    detector_backend = "yolov8",
    align = True,
    expand_percentage = 0,
    threshold = 0.90,
    normalization = "base",
    silent = False,
)

24-05-10 06:31:52 - Searching assets/Eiza_Gonzalez_03.jpg in 27 length datastore 24-05-10 06:31:53 - find function duration 0.1595611572265625 seconds

#	identity	hash	target_x	target_y	target_w	target_h	source_x	source_y	source_w	source_h	threshold	distance
0	assets/Eiza_Gonzalez_03.jpg	c851bc623b011079548c899eedc96a0b8c1a9264	39	19	62	84	39	19	62	84	0.9	-2.220446e-16
1	assets/Eiza_Gonzalez_01.jpg	dd70311d238d2be8b47aee363e3f228b741c3b50	203	65	139	183	39	19	62	84	0.9	1.577659e-01
2	assets/Eiza_Gonzalez_05.jpg	4e3c950ef9084dd264e056711afd7a949dfb5f95	57	53	119	163	39	19	62	84	0.9	3.097999e-01
3	assets/Eiza_Gonzalez_04.jpg	863a9e1f9f06ff1da45ef39de64f03a24d768164	19	23	79	97	39	19	62	84	0.9	4.673658e-01
4	assets/Eiza_Gonzalez_02.jpg	220a99ec5380b5d4033dd18f6f54353a49bb341b	497	132	192	252	39	19	62	84	0.9	7.133165e-01
5	assets/Benedict_Wong_05.jpg	4f14065c9065e745f63b1ca22c206649922eaee2	45	18	36	49	39	19	62	84	0.9	8.396030e-01
6	assets/Liam_Cunningham_01.jpg	a157a9143b88b9844ea65f1ccea484a337670853	137	163	231	319	39	19	62	84	0.9	8.708137e-01
7	assets/Jovan_Adepo_01.jpg	47bc06e224a2538aa258c0b5d03f478e5713b8bc	46	16	64	92	39	19	62	84	0.9	8.942636e-01
8	assets/Liam_Cunningham_03.jpg	d349c11690daa1a9e118df5d4ae08d554852a96a	118	58	130	184	39	19	62	84	0.9	8.980692e-01

image_results = [
    ['./assets/Eiza_Gonzalez_03.jpg',  1.0000000000000002],
    ['./assets/Eiza_Gonzalez_01.jpg',  0.8422341],
    ['./assets/Eiza_Gonzalez_05.jpg',  0.6902001],
    ['./assets/Eiza_Gonzalez_04.jpg',  0.5326342],
    ['./assets/Eiza_Gonzalez_02.jpg',  0.2866835],
    ['./assets/Benedict_Wong_05.jpg',  0.160397]
]

plt.figure(figsize=(12,8))

for i in range(len(image_results)):
    ax = plt.subplot(2,3,i+1)
    plt.title(
        image_results[i][0][9:-4]+'\n Similarity: '+str(round(image_results[i][1],2))
    )
    img = plt.imread(image_results[i][0])
    plt.imshow(img)
    plt.axis('off')

data_images = [
    'Eiza_Gonzalez_03.jpg',
    'Eiza_Gonzalez_01.jpg',
    'Eiza_Gonzalez_05.jpg',
    'Eiza_Gonzalez_04.jpg',
    'Eiza_Gonzalez_02.jpg',
    'Benedict_Wong_05.jpg',
    'Liam_Cunningham_01.jpg',
    'Jovan_Adepo_01.jpg',
    'Liam_Cunningham_03.jpg'
]

distance_data = [
     -2.220446e-16,
     1.577659e-01,
     3.097999e-01,
     4.673658e-01,
     7.133165e-01,
     8.396030e-01,
     8.708137e-01,
     8.942636e-01,
     8.980692e-01
]

similarity_data = [
     1.0000000000000002,
     0.8422341,
     0.6902001,
     0.5326342,
     0.2866835,
     0.160397,
     0.12918629999999998,
     0.10573639999999995,
     0.10193079999999999
]

result_10_df = pd.DataFrame({ 
    'Images': data_images, 
    'Distance': distance_data, 
    'Similarity': similarity_data  
}) 

result_10_df

	Images	Distance	Similarity
0	Eiza_Gonzalez_03.jpg	-2.220446e-16	1.000000
1	Eiza_Gonzalez_01.jpg	1.577659e-01	0.842234
2	Eiza_Gonzalez_05.jpg	3.097999e-01	0.690200
3	Eiza_Gonzalez_04.jpg	4.673658e-01	0.532634
4	Eiza_Gonzalez_02.jpg	7.133165e-01	0.286683
5	Benedict_Wong_05.jpg	8.396030e-01	0.160397
6	Liam_Cunningham_01.jpg	8.708137e-01	0.129186
7	Jovan_Adepo_01.jpg	8.942636e-01	0.105736
8	Liam_Cunningham_03.jpg	8.980692e-01	0.101931

result_10_df.plot(x="Images", y="Similarity", kind="bar")

ArcFace / YOLOv8

• arcface_weights.h5 - 130.7 MB
• yolov8n-face.pt - 6.1 MB
• 24-05-10 06:51:33 - Searching assets/Eiza_Gonzalez_03.jpg in 27 length datastore
• 24-05-10 06:51:34 - Find function duration 0.14449524879455566 seconds

result11 = DeepFace.find(
    img_path = "assets/Eiza_Gonzalez_03.jpg",
    db_path = "assets",
    model_name = "ArcFace",
    distance_metric = "cosine",
    enforce_detection = True,
    detector_backend = "yolov8",
    align = True,
    expand_percentage = 0,
    threshold = 0.90,
    normalization = "base",
    silent = False,
)

24-05-10 06:59:27 - Searching assets/Eiza_Gonzalez_03.jpg in 27 length datastore 24-05-10 06:59:27 - find function duration 0.14449524879455566 seconds

#	identity	hash	target_x	target_y	target_w	target_h	source_x	source_y	source_w	source_h	threshold	distance
0	assets/Eiza_Gonzalez_03.jpg	c851bc623b011079548c899eedc96a0b8c1a9264	39	19	62	84	39	19	62	84	0.9	0.000000
1	assets/Eiza_Gonzalez_01.jpg	dd70311d238d2be8b47aee363e3f228b741c3b50	203	65	139	183	39	19	62	84	0.9	0.491314
2	assets/Eiza_Gonzalez_04.jpg	863a9e1f9f06ff1da45ef39de64f03a24d768164	19	23	79	97	39	19	62	84	0.9	0.534116
3	assets/Eiza_Gonzalez_05.jpg	4e3c950ef9084dd264e056711afd7a949dfb5f95	57	53	119	163	39	19	62	84	0.9	0.597161
4	assets/Eiza_Gonzalez_02.jpg	220a99ec5380b5d4033dd18f6f54353a49bb341b	497	132	192	252	39	19	62	84	0.9	0.784296
5	assets/Alex_Sharp_03.jpg	48a617ed38787e3dd5e7d7e87e2a56b37214ce38	110	111	248	345	39	19	62	84	0.9	0.828761
6	assets/Jovan_Adepo_01.jpg	47bc06e224a2538aa258c0b5d03f478e5713b8bc	46	16	64	92	39	19	62	84	0.9	0.858612
7	assets/Jovan_Adepo_02.jpg	f5428170e3323a333121ceb196675526d8ad02c6	63	59	117	169	39	19	62	84	0.9	0.866306
8	assets/Alex_Sharp_01.jpg	7155ec0a507fd98583c179beb164c9fcfa0fa917	207	42	93	123	39	19	62	84	0.9	0.887515

image_results = [
    ['./assets/Eiza_Gonzalez_03.jpg',  1.000000000000000],
    ['./assets/Eiza_Gonzalez_01.jpg',  0.508686],
    ['./assets/Eiza_Gonzalez_04.jpg',  0.46588399999999996],
    ['./assets/Eiza_Gonzalez_05.jpg',  0.40283899999999995],
    ['./assets/Eiza_Gonzalez_02.jpg',  0.215704],
    ['./assets/Alex_Sharp_03.jpg',  0.17123900000000003]
]

plt.figure(figsize=(12,8))

for i in range(len(image_results)):
    ax = plt.subplot(2,3,i+1)
    plt.title(
        image_results[i][0][9:-4]+'\n Similarity: '+str(round(image_results[i][1],2))
    )
    img = plt.imread(image_results[i][0])
    plt.imshow(img)
    plt.axis('off')

data_images = [
    'Eiza_Gonzalez_03.jpg',
    'Eiza_Gonzalez_01.jpg',
    'Eiza_Gonzalez_04.jpg',
    'Eiza_Gonzalez_05.jpg',
    'Eiza_Gonzalez_02.jpg',
    'Alex_Sharp_03.jpg',
    'Jovan_Adepo_01.jpg',
    'Jovan_Adepo_02.jpg',
    'Alex_Sharp_01.jpg'
]

distance_data = [
     0.000000,
     0.491314,
     0.534116,
     0.597161,
     0.784296,
     0.828761,
     0.858612,
     0.866306,
     0.887515
]

similarity_data = [
     1.0,
     0.508686,
     0.46588399999999996,
     0.40283899999999995,
     0.215704,
     0.17123900000000003,
     0.14138799999999996,
     0.13369399999999998,
     0.11248499999999995
]

result_11_df = pd.DataFrame({ 
    'Images': data_images, 
    'Distance': distance_data, 
    'Similarity': similarity_data  
}) 

result_11_df

	Images	Distance	Similarity
0	Eiza_Gonzalez_03.jpg	0.000000	1.000000
1	Eiza_Gonzalez_01.jpg	0.491314	0.508686
2	Eiza_Gonzalez_04.jpg	0.534116	0.465884
3	Eiza_Gonzalez_05.jpg	0.597161	0.402839
4	Eiza_Gonzalez_02.jpg	0.784296	0.215704
5	Alex_Sharp_03.jpg	0.828761	0.171239
6	Jovan_Adepo_01.jpg	0.858612	0.141388
7	Jovan_Adepo_02.jpg	0.866306	0.133694
8	Alex_Sharp_01.jpg	0.887515	0.112485

result_11_df.plot(x="Images", y="Similarity", kind="bar")

GhostFaceNet / YOLOv8

• GhostFaceNet_W1.3_S1_ArcFace.h5 - 16.5 MB
• yolov8n-face.pt - 6.1 MB
• 24-05-10 07:47:47 - Searching assets/Eiza_Gonzalez_03.jpg in 27 length datastore
• 24-05-10 07:47:47 - Find function duration 0.16495823860168457 seconds

result12 = DeepFace.find(
    img_path = "assets/Eiza_Gonzalez_03.jpg",
    db_path = "assets",
    model_name = "GhostFaceNet",
    distance_metric = "cosine",
    enforce_detection = True,
    detector_backend = "yolov8",
    align = True,
    expand_percentage = 0,
    threshold = 0.90,
    normalization = "base",
    silent = False,
)

24-05-10 07:47:47 - Searching assets/Eiza_Gonzalez_03.jpg in 27 length datastore 24-05-10 07:47:47 - find function duration 0.16495823860168457 seconds

#	identity	hash	target_x	target_y	target_w	target_h	source_x	source_y	source_w	source_h	threshold	distance
0	assets/Eiza_Gonzalez_03.jpg	c851bc623b011079548c899eedc96a0b8c1a9264	39	19	62	84	39	19	62	84	0.9	0.000000
1	assets/Eiza_Gonzalez_01.jpg	dd70311d238d2be8b47aee363e3f228b741c3b50	203	65	139	183	39	19	62	84	0.9	0.588340
2	assets/Eiza_Gonzalez_05.jpg	4e3c950ef9084dd264e056711afd7a949dfb5f95	57	53	119	163	39	19	62	84	0.9	0.713904
3	assets/Jess_Hong_02.jpg	1233412de754501680ab281fdd0b3c0b7e634279	43	14	67	89	39	19	62	84	0.9	0.799567
4	assets/Eiza_Gonzalez_04.jpg	863a9e1f9f06ff1da45ef39de64f03a24d768164	19	23	79	97	39	19	62	84	0.9	0.814843
5	assets/Jess_Hong_03.jpg	9b056b12303c62eeee91c49456bdda7f494aa078	206	24	101	135	39	19	62	84	0.9	0.857111
6	assets/Benedict_Wong_05.jpg	4f14065c9065e745f63b1ca22c206649922eaee2	45	18	36	49	39	19	62	84	0.9	0.861572
7	assets/Eiza_Gonzalez_02.jpg	220a99ec5380b5d4033dd18f6f54353a49bb341b	497	132	192	252	39	19	62	84	0.9	0.890497

image_results = [
    ['./assets/Eiza_Gonzalez_03.jpg',  1.000000000000000],
    ['./assets/Eiza_Gonzalez_01.jpg',  0.41166],
    ['./assets/Eiza_Gonzalez_05.jpg',  0.286096],
    ['./assets/Jess_Hong_02.jpg',  0.20043299999999997],
    ['./assets/Eiza_Gonzalez_04.jpg',  0.18515700000000002],
    ['./assets/Jess_Hong_03.jpg',  0.14288900000000004]
]

plt.figure(figsize=(12,8))

for i in range(len(image_results)):
    ax = plt.subplot(2,3,i+1)
    plt.title(
        image_results[i][0][9:-4]+'\n Similarity: '+str(round(image_results[i][1],2))
    )
    img = plt.imread(image_results[i][0])
    plt.imshow(img)
    plt.axis('off')

data_images = [
    'Eiza_Gonzalez_03.jpg',
    'Eiza_Gonzalez_01.jpg',
    'Eiza_Gonzalez_05.jpg',
    'Jess_Hong_02.jpg',
    'Eiza_Gonzalez_04.jpg',
    'Jess_Hong_03.jpg',
    'Benedict_Wong_05.jpg',
    'Eiza_Gonzalez_02.jpg'
]

distance_data = [
     0.000000,
     0.588340,
     0.713904,
     0.799567,
     0.814843,
     0.857111,
     0.861572,
     0.890497
]

similarity_data = [
     1.0,
     0.41166,
     0.286096,
     0.20043299999999997,
     0.18515700000000002,
     0.14288900000000004,
     0.138428,
     0.10950300000000002
]

result_12_df = pd.DataFrame({ 
    'Images': data_images, 
    'Distance': distance_data, 
    'Similarity': similarity_data  
}) 

result_12_df

	Images	Distance	Similarity
0	Eiza_Gonzalez_03.jpg	0.000000	1.000000
1	Eiza_Gonzalez_01.jpg	0.588340	0.411660
2	Eiza_Gonzalez_05.jpg	0.713904	0.286096
3	Jess_Hong_02.jpg	0.799567	0.200433
4	Eiza_Gonzalez_04.jpg	0.814843	0.185157
5	Jess_Hong_03.jpg	0.857111	0.142889
6	Benedict_Wong_05.jpg	0.861572	0.138428
7	Eiza_Gonzalez_02.jpg	0.890497	0.109503

result_12_df.plot(x="Images", y="Similarity", kind="bar")

Face Analysis

• facial_expression_model_weights.h5 - 5.7 MB
• age_model_weights.h5 - 513.8 MB
• gender_model_weights.h5 - 512.3 MB
• race_model_single_batch.h5 - 512.3 MB

result13 = DeepFace.analyze(
    img_path = "assets/Benedict_Wong_03.jpg", 
    actions = ["emotion", "age", "gender", "race"],
    enforce_detection = True,
    detector_backend = "yolov8",
    align = True,
    expand_percentage = 0,
    silent = False,
)

pretty_print2(result13)

Emotions: angry 52.47458815574646, disgust 1.6047310500973637e-14, fear 1.1184778797579398e-08, happy 0.8838735520839691, sad 0.002243508606625255, surprise 1.661406301423085e-06, neutral 46.63929641246796 Dominant Emotion: angry, Confidence: 0.87, Age: 44, Female: 0.08514390792697668, Male: 99.91486072540283, Dominant Gender: Man, Asian: 99.96744990154802, Indian: 0.020422336582121944, Black: 3.538733961399402e-07, White: 0.0007214643549394969, Middle Eastern: 9.568194723961802e-09, Latino Hispanic: 0.011405441862250006, Dominant Race: asian, Detections :: x,y,w,h,left_eye,right_eye Regions: 89,84,130,173,(183, 149),(125, 146)

result14 = DeepFace.analyze(
    img_path = "assets/Eiza_Gonzalez_03.jpg", 
    actions = ["emotion", "age", "gender", "race"],
    enforce_detection = True,
    detector_backend = "yolov8",
    align = True,
    expand_percentage = 0,
    silent = False,
)

pretty_print2(result14)

Emotions: angry 61.89677715301514, disgust 0.0003831427875411464, fear 17.56562739610672, happy 0.024294608738273382, sad 7.415241003036499, surprise 0.0922270177397877, neutral 13.00545483827591 Dominant Emotion: angry, Confidence: 0.83, Age: 27, Female: 99.99103546142578, Male: 0.008965710730990395, Dominant Gender: Woman, Asian: 1.2231661533613387, Indian: 1.5839036579613277, Black: 0.15020265171857497, White: 51.279270801457265, Middle Eastern: 18.619856967908973, Latino Hispanic: 27.143603818846206, Dominant Race: white, Detections :: x,y,w,h,left_eye,right_eye Regions: 39,19,62,84,(91, 57),(66, 52)

result15 = DeepFace.analyze(
    img_path = "assets/Jovan_Adepo_01.jpg", 
    actions = ["emotion", "age", "gender", "race"],
    enforce_detection = True,
    detector_backend = "yolov8",
    align = True,
    expand_percentage = 0,
    silent = False,
)

pretty_print2(result15)

Emotions: angry 0.043431183439679444, disgust 1.5149926940131314e-11, fear 0.0003094855173912947, happy 4.4355454065225786e-05, sad 0.051101919962093234, surprise 2.526942033398427e-06, neutral 99.90511536598206 Dominant Emotion: neutral, Confidence: 0.86, Age: 35, Female: 0.016746165056247264, Male: 99.98325109481812, Dominant Gender: Man, Asian: 1.4956607063965066e-07, Indian: 4.022327360075906e-06, Black: 100.0, White: 2.5537931144792303e-10, Middle Eastern: 9.079970391019654e-11, Latino Hispanic: 1.2836626872569923e-06, Dominant Race: black, Detections :: x,y,w,h,left_eye,right_eye Regions: 46,16,64,92,(92, 51),(61, 50)

result16 = DeepFace.analyze(
    img_path = "assets/Jess_Hong_01.jpg", 
    actions = ["emotion", "age", "gender", "race"],
    enforce_detection = True,
    detector_backend = "yolov8",
    align = True,
    expand_percentage = 0,
    silent = False,
)

pretty_print2(result16)

Emotions: angry 21.61679118871689, disgust 0.002775121174636297, fear 8.90427678823471, happy 0.0006556400876434054, sad 53.71009111404419, surprise 0.0014677107174065895, neutral 15.763948857784271 Dominant Emotion: sad, Confidence: 0.85, Age: 35, Female: 96.56398296356201, Male: 3.4360110759735107, Dominant Gender: Woman, Asian: 96.5178076603642, Indian: 1.5434301481032195, Black: 0.11495926676803628, White: 0.3914411644409699, Middle Eastern: 0.02275169119265259, Latino Hispanic: 1.4096110310124121, Dominant Race: asian, Detections :: x,y,w,h,left_eye,right_eye Regions: 110,54,48,69,(132, 80),(114, 80)

models = [
    'VGG-Face / OpenCV',
    'Dlib / dlib',
    'Dlib / Mediapipe',
    'Dlib / YOLOv8',
    'VGG-Face / YOLOv8',
    'Facenet512 / YOLOv8',
    'Facenet / YOLOv8',
    'ArcFace / YOLOv8',
    'GhostFaceNet / YOLOv8'
]

weight_size = [
     552.3,
     20.4,
     25.8,
     26.1,
     558.4,
     96.7,
     94.0,
     136.8,
     22.6
]

run_time = [
     0.21913480758666992,
     0.13037323951721191,
     0.3898661136627197,
     0.12443137168884277,
     0.21353983879089355,
     0.1690361499786377,
     0.1595611572265625,
     0.14449524879455566,
     0.16495823860168457
]

models_df = pd.DataFrame({ 
    'Face Recognition / Detection': models, 
    'Combined Model Weight Size [MB]': weight_size, 
    'Detection Time [s]': run_time  
}) 

models_df

	Face Recognition / Detection	Combined Model Weight Size [MB]	Detection Time [s]
0	VGG-Face / OpenCV	552.3	0.219135
1	Dlib / dlib	20.4	0.130373
2	Dlib / Mediapipe	25.8	0.389866
3	Dlib / YOLOv8	26.1	0.124431
4	VGG-Face / YOLOv8	558.4	0.213540
5	Facenet512 / YOLOv8	96.7	0.169036
6	Facenet / YOLOv8	94.0	0.159561
7	ArcFace / YOLOv8	136.8	0.144495
8	GhostFaceNet / YOLOv8	22.6	0.164958

models_df.plot(x="Face Recognition / Detection", y="Combined Model Weight Size [MB]", kind="bar")

models_df.plot(x="Face Recognition / Detection", y="Detection Time [s]", kind="bar")