Vision Analysis
All Hardware

NVIDIA A100

NVIDIA A100-PCIE-40GB

VRAM

40 GB

FP32 Performance

19.5 TFLOPS

Power (TDP)

250W

Highest Accuracy
Top 5 models by mAP@50-95
1.
dfine-x
59.3%
2.
dfine-l
57.3%
3.
dfine-m
55.1%
4.
rtdetr-r101
53.9%
5.
rtdetr-r50
52.7%
Fastest
Top 5 models by throughput
1.
rtdetr-r34
28 FPS
2.
rtdetr-r18
27 FPS
3.
dfine-n
26 FPS
4.
rtdetr-r50m
25 FPS
5.
dfine-s
24 FPS
All Models
13 models benchmarked on NVIDIA A100 (PyTorch FP32)
#ModelmAPFPSLatencyParams
1
dfine-x		59.3%15.464.7ms62.6M
2
dfine-l		57.3%17.158.3ms31.2M
3
dfine-m		55.1%21.746.1ms19.6M
4
rtdetr-r101		53.9%18.653.6ms76.6M
5
rtdetr-r50		52.7%22.744.1ms42.9M
6
yolonas-l		51.2%16.660.4ms67.0M
7
rtdetr-r50m		50.8%25.139.8ms36.6M
8
dfine-s		50.7%24.041.7ms10.3M
9
yolonas-m		50.5%17.058.9ms51.2M
10
rtdetr-r34		48.2%27.836.0ms31.4M
11
yolonas-s		46.5%16.660.1ms19.1M
12
rtdetr-r18		45.6%27.436.5ms20.2M
13
dfine-n		42.8%25.639.1ms3.8M

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