ABCDEFGHIJKLMNOPQRSTUVWXYZ
1
Description - https://medium.com/@zlodeibaal/choosing-computer-vision-board-in-2022-b27eb4ca7a7c
CORALKHADAS VIM3ESP32Raspberry Pi 3BIntel NCS2Rock Pi 3A 3568Jetson NanoHailo8MAIX-III AX620A RK3588RK3566Raspberry Pi 5
Intel i5 i5-6600 OpenVino
2
System itself
3
Easy to flash45354.54.54.54.54.54.54.55
4
Easy to work with5334.553.54.542.53.53.54.5
5
Conventional Linux4.54.50554.54.5- 34.54.55
6
Community support434553512335
7
Models
8
Oficial Models Zoo52.504.553554.5334.5
9
Unofficial Models Zoo4.51.51552511225
10
Easy to export random model3.54*25544.554445
11
Easy to debug convertation44.5*0553553.5335
12
Hardware
13
Processor speed3.5414.503.53.5- 4.543.54.5
14
test 1****18247518501332571015812
15
test 2****1918348831449938267482367633043877664
16
17
mechanical and construction4444.54545/32554.5
18
How easy is it to buy?555554435445
19
Pins for external connection553505505555
20
Models (Batch 1)TFlite TPUKhadasbinary:)ONNXtflite fp16 cv2-ONNXNCNNTorchScript int8TorchScript fp32FP16-32OAKrknn-toolkit
TensorRT FP32/16
int8int8rknn-toolkitNCNNrknn-toolkittflite fp16 ONNXNCNNINT8FP16FP32
21
MobileNet v1 224*2242.9ms (int8)7ms**(int8)360-46024ms<33ms9(int8)20ms319(int8)3.468
22
MobileNet v2 224*2243ms(int8)5ms***(int8)210-410ms160-300240-300ms100130030ms10(int8)20ms3.53.326(int8)32-55ms14ms557
23
MobileNet v1 a=0.25 96*96900ms(int8)
24
SSD MobileNet v1 300*300 90class
10ms(int8)700ms350-430ms'41ms<33ms30ms101113
25
SSD mobileNet v2 300*300 90class
15ms(int8)380-450ms'62ms<74ms38ms71010
26
SSD mobileNet v1 320*320 90class18ms**(int8)
27
Yolov5S
28
640*640/NoNMS200ms(int8)100ms***(int8)3000-4800ms3400-11000ms5900-6300ms3000-3300ms'340ms<200ms170(int8)120''202212ms (fp16)85ms330 (int8)650ms450ms165-210ms82
29
224*224/NoNMS47ms(int8)30ms***(int8)400-780ms430-550ms730-920ms490-550ms'50ms<33ms18(int8) 50(fp16)305110 (fp16)70ms37-63ms21ms13
30
ResNet5051.5ms(int8)27ms(int8)1300-2300ms2300-2450ms170063ms<66ms44(int8)292822.6110(int8)135-175ms67-89ms213333
31
Other
32
Price (USD)130120535-50 (on 09.2020 the price is around 131)10035270'''~40-120 Hailo~60-120
33
Project I was involver in010511>1000
34
Hobby project I made001210100
35
Link
https://www.youtube.com/watch?v=hMnzSk35lSU
https://www.youtube.com/watch?v=ii-KpD-KdYw
https://youtu.be/w_sCTPDuutQ
https://youtu.be/EE2x3BhxubAhttps://youtu.be/ATaYQ3FBFxk
https://youtu.be/Sadmw6Rrj1Y
https://youtu.be/ja7HPgJWcg4
https://youtu.be/kfnz8kB2FLA
https://youtu.be/3q89nC_tHE0
36
Idle power consumption2.2w2.6w1.65w0.85w1.65w2.2w
37
Neural network inference4.35w3.5w2.05w1.75w2.2w>10w2w hailo + board3w
38
* - Dissenting opinion, there is ambiguity
' - models from examples (not mine convertion)
*Last update from 2024. Everything easy now
39
** - converted from TFlite
" - on Jetson Nano NMS is VERY slow.
40
*** - converted from ONNX
"' - few years ago the price was ~100usd
41
**** - stress-ng --cpu 4 --cpu-method matrixprod --metrics-brief --perf -t 20
42
**** - stress-ng --cpu 4 --cpu-method factorial --metrics-brief --perf -t 20
43
Link to description - https://medium.com/@zlodeibaal/choosing-computer-vision-board-in-2022-b27eb4ca7a7c
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100