Vision Analysis
All Hardware

Raspberry Pi 5

Broadcom BCM2712

RAM

16 GB

Power (TDP)

12W

Runtime:
Highest Accuracy
Top 5 models by mAP@50-95
61.3%
61.1%
60.1%
60.0%
58.6%
Fastest
Top 5 models by throughput
13 FPS
11 FPS
9 FPS
5 FPS
All Models
70 models benchmarked on Raspberry Pi 5 (PyTorch FP32)
#ModelmAPFPSLatencyParams
1
deimv2-x
61.3%0.33711.8ms51.2M
2
ec-x
61.1%0.33635.1ms49.9M
3
ec-l
60.1%0.32850.3ms33.0M
4
dfine-l
60.0%0.42305.2ms31.2M
5
rfdetr-l
58.6%0.52110.2ms33.9M
6
deimv2-l
58.6%0.42802.2ms32.5M
7
ec-m
58.4%0.51868.7ms19.4M
8
deim-l
57.8%0.42295.5ms31.2M
9
dfine-m
57.8%0.61577.2ms19.6M
10
rtdetrv4-l
57.8%0.42303.8ms31.2M
11
rfdetr-m
57.4%0.81281.0ms33.7M
12
rtdetrv4-m
56.5%0.61588.4ms19.6M
13
yolo26x
56.5%0.42507.6ms55.7M
14
yolov9c
56.5%0.52153.8ms25.5M
15
deimv2-m
56.0%0.51913.2ms18.4M
16
rtdetr-r50
55.9%0.33317.6ms42.9M
17
rtdetr-l
55.8%0.42345.2ms32.9M
18
rtdetrv2-r50
55.7%0.33317.3ms42.9M
19
yolonas-m
55.5%0.52028.6ms51.2M
20
deim-m
55.4%0.61582.8ms19.6M
21
yolov9m
55.3%0.61638.0ms20.1M
22
rfdetr-s
55.1%1.0959.4ms32.1M
23
rtdetrv2-r50m
54.8%0.42810.3ms36.6M
24
ec-s
54.3%0.81201.0ms9.9M
25
yolox-l
53.9%0.42627.0ms54.2M
26
rtdetr-r50m
53.8%0.42802.0ms36.6M
27
yolo26l
53.8%0.71377.3ms24.8M
28
yolo11x
53.6%0.42522.8ms56.9M
29
dfine-s
53.4%1.1874.3ms10.3M
30
rtdetrv2-r34
53.2%0.61664.9ms31.4M
31
yolov8x
53.0%0.42422.0ms68.2M
32
deimv2-s
53.0%0.81190.9ms9.8M
33
rtdetrv4-s
52.8%1.1879.9ms10.3M
34
yolov5xu
52.3%0.42484.4ms97.2M
35
yolo11l
52.3%0.71385.3ms25.3M
36
rtdetr-r34
52.2%0.61669.0ms31.4M
37
yolo26m
52.2%0.91108.3ms20.4M
38
deim-s
52.1%1.1878.2ms10.3M
39
yolov8l
52.0%0.61762.9ms43.7M
40
yolonas-s
51.8%0.91049.5ms19.1M
41
yolov5lu
51.5%0.61587.0ms53.2M
42
rfdetr-n
51.4%2.1479.6ms30.5M
43
yolox-m
50.9%0.81284.1ms25.3M
44
rtdetrv2-r18
50.8%0.81244.5ms20.2M
45
yolo11m
50.6%0.91104.3ms20.1M
46
rtdetr-r18
49.7%0.81245.4ms20.2M
47
yolov9s
49.5%1.4716.7ms7.2M
48
yolov8m
49.4%1.1906.5ms25.9M
49
yolov5mu
48.2%1.2824.2ms25.1M
50
yolo26s
47.4%2.3441.8ms9.5M
51
deim-n
46.8%2.3431.0ms3.8M
52
deimv2-n
46.7%2.4423.0ms3.6M
53
dfine-n
45.8%2.3429.1ms3.8M
54
yolo11s
45.7%2.3435.8ms9.4M
55
yolov8s
44.2%2.4421.1ms11.2M
56
picodet-l
44.2%1.9540.3ms3.3M
57
yolox-s
43.0%1.6626.6ms9.0M
58
yolov5su
42.4%2.5397.4ms9.1M
59
deimv2-pico
42.2%2.9349.8ms1.5M
60
yolov9t
41.4%2.9346.3ms2.0M
61
yolo26n
40.1%5.0198.1ms2.4M
62
yolo11n
38.7%4.9204.7ms2.6M
63
picodet-m
37.6%5.0201.1ms2.1M
64
yolov8n
36.7%5.3186.8ms3.1M
65
yolox-tiny
35.1%5.5183.5ms5.1M
66
deimv2-femto
34.5%7.4135.3ms1.0M
67
yolov5nu
33.8%5.4184.6ms2.6M
68
picodet-s
29.5%10.595.0ms1.0M
69
yolox-nano
28.8%8.6117.0ms0.9M
70
deimv2-atto
27.5%12.779.1ms0.5M

Run any model with one line

LibreYOLO has the best catalogue of state-of-the-art detectors, all behind one MIT-licensed Python API.

from libreyolo import LibreYOLO, SAMPLE_IMAGE

# LibreYOLO has the best catalogue of state-of-the-art models.
model = LibreYOLO("LibreRFDETRl.pt")           # RF-DETR-L (transformer flagship)
results = model(SAMPLE_IMAGE, save=True)        # run inference, save the annotated image

# Swap in any other model, same one-line API (weights auto-download):
#   LibreYOLO("LibreYOLO9c.pt")      # YOLO9-C
#   LibreYOLO("LibreYOLOXx.pt")      # YOLOX-X
#   LibreYOLO("LibreDFINEx.pt")      # D-FINE-X
#   LibreYOLO("LibreRTDETRr50.pt")   # RT-DETR-R50