Object Detection Leaderboard
Understand the computer vision landscape to select the best model and hardware for your use case.
Accuracy
mAP@50-95 scores on COCO val2017 - the standard benchmark for object detection accuracy.
54 models
Accuracy vs Model Size
mAP@50-95 on COCO val2017 plotted against parameter count. Higher and left is better.
damoyolo
deim
deimv2
dfine
ec
rfdetr
rtdetr
rtdetrv2
rtdetrv4
yolox
Leaderboard
Full ranking of all benchmarked models on NVIDIA RTX 5070 Ti (PYTORCH FP32). Speed numbers reflect this hardware — switch hardware in the filter above to compare. Click column headers to sort.
| # | Model | mAP@50-95 | mAP@50 | FPS | Latency | Params (M) | GFLOPs | mAP/GFLOP | |
|---|---|---|---|---|---|---|---|---|---|
1 | dfine-x | 59.3% | 76.8% | 18.9 | 52.8ms | 62.6M | 202.0 | 0.3 | Model |
2 | ec-x | 57.9% | 76.0% | 26.5 | 37.8ms | 49.9M | 151.0 | 0.4 | Model |
3 | deimv2-x | 57.8% | 75.3% | 19.6 | 51.0ms | 51.2M | 151.6 | 0.4 | Model |
4 | dfine-l | 57.3% | 74.9% | 25.5 | 39.2ms | 31.2M | 91.0 | 0.6 | Model |
5 | rtdetrv4-x | 57.0% | 74.6% | 21.9 | 45.8ms | 62.6M | 202.0 | 0.3 | Model |
6 | ec-l | 57.0% | 75.1% | 27.1 | 36.9ms | 33.0M | 101.0 | 0.6 | Model |
7 | rfdetr-l | 56.5% | 75.1% | 33.3 | 30.0ms | 33.9M | 340.0 | 0.2 | Model |
8 | deim-x | 56.5% | 74.0% | 22.7 | 44.0ms | 62.6M | 202.0 | 0.3 | Model |
9 | deimv2-l | 56.0% | 73.5% | 22.2 | 45.0ms | 32.5M | 96.3 | 0.6 | Model |
10 | rtdetrv4-l | 55.4% | 73.1% | 25.0 | 40.0ms | 31.2M | 91.0 | 0.6 | Model |
11 | dfine-m | 55.1% | 72.6% | 34.2 | 29.2ms | 19.6M | 57.0 | 1.0 | Model |
12 | deim-l | 54.7% | 72.4% | 27.2 | 36.7ms | 31.2M | 91.0 | 0.6 | Model |
13 | rfdetr-m | 54.7% | 73.5% | 39.7 | 25.2ms | 33.7M | 0.0 | 0.0 | Model |
14 | rtdetr-x | 54.7% | 72.9% | 23.7 | 42.1ms | 67.4M | 234.0 | 0.2 | Model |
15 | rtdetr-r101 | 54.4% | 72.8% | 25.3 | 39.5ms | 76.6M | 259.0 | 0.2 | Model |
16 | rtdetrv2-r101 | 54.4% | 72.8% | 22.4 | 44.5ms | 76.6M | 259.0 | 0.2 | Model |
17 | ec-m | 54.2% | 72.2% | 31.4 | 31.8ms | 19.4M | 53.0 | 1.0 | Model |
18 | rtdetrv4-m | 53.6% | 71.0% | 35.1 | 28.5ms | 19.6M | 57.0 | 0.9 | Model |
19 | rtdetrv2-r50 | 53.4% | 71.6% | 30.4 | 32.9ms | 42.9M | 136.0 | 0.4 | Model |
20 | rtdetr-r50 | 53.1% | 71.2% | 32.9 | 30.4ms | 42.9M | 136.0 | 0.4 | Model |
21 | deimv2-m | 53.0% | 70.3% | 23.9 | 41.8ms | 18.4M | 52.2 | 1.0 | Model |
22 | rfdetr-s | 53.0% | 72.1% | 47.0 | 21.3ms | 32.1M | 0.0 | 0.0 | Model |
23 | rtdetr-l | 52.9% | 71.5% | 28.9 | 34.6ms | 32.9M | 110.0 | 0.5 | Model |
24 | deim-m | 52.7% | 70.0% | 35.3 | 28.3ms | 19.6M | 57.0 | 0.9 | Model |
25 | rtdetrv2-r50m | 51.9% | 69.8% | 35.5 | 28.1ms | 36.6M | 100.0 | 0.5 | Model |
26 | ec-s | 51.7% | 69.4% | 32.0 | 31.3ms | 9.9M | 26.0 | 2.0 | Model |
27 | rtdetr-r50m | 51.3% | 69.5% | 37.7 | 26.5ms | 36.6M | 0.0 | 0.0 | Model |
28 | yolox-x | 51.0% | 69.2% | 30.7 | 32.6ms | 99.1M | 141.2 | 0.4 | Model |
29 | deimv2-s | 50.9% | 68.3% | 27.3 | 36.6ms | 9.8M | 25.6 | 2.0 | Model |
30 | dfine-s | 50.7% | 67.7% | 35.1 | 28.5ms | 10.3M | 25.0 | 2.0 | Model |
31 | damoyolo-m | 50.0% | 67.5% | 8.7 | 114.9ms | 28.2M | 61.8 | 0.8 | Model |
32 | rtdetrv2-r34 | 49.9% | 67.6% | 52.9 | 18.9ms | 31.4M | 92.0 | 0.5 | Model |
33 | rtdetrv4-s | 49.8% | 67.1% | 45.1 | 22.2ms | 10.3M | 25.0 | 2.0 | Model |
34 | yolox-l | 49.6% | 68.1% | 38.9 | 25.7ms | 54.2M | 78.0 | 0.6 | Model |
35 | deim-s | 49.0% | 65.9% | 43.8 | 22.8ms | 10.3M | 25.0 | 2.0 | Model |
36 | rtdetr-r34 | 48.9% | 66.7% | 44.5 | 22.5ms | 31.4M | 91.0 | 0.5 | Model |
37 | rfdetr-n | 48.4% | 67.5% | 55.8 | 17.9ms | 30.5M | 0.0 | 0.0 | Model |
38 | rtdetrv2-r18 | 48.1% | 65.1% | 62.1 | 16.1ms | 20.2M | 60.0 | 0.8 | Model |
39 | yolox-m | 46.8% | 65.7% | 41.7 | 24.0ms | 25.3M | 37.0 | 1.3 | Model |
40 | rtdetr-r18 | 46.4% | 63.7% | 51.4 | 19.4ms | 20.2M | 60.0 | 0.8 | Model |
41 | damoyolo-s | 45.9% | 62.9% | 8.4 | 118.6ms | 16.3M | 37.8 | 1.2 | Model |
42 | deim-n | 43.0% | 60.4% | 49.5 | 20.2ms | 3.8M | 7.0 | 6.2 | Model |
43 | deimv2-n | 43.0% | 60.1% | 38.8 | 25.8ms | 3.6M | 6.9 | 6.3 | Model |
44 | dfine-n | 42.8% | 60.3% | 39.6 | 25.3ms | 3.8M | 7.0 | 6.1 | Model |
45 | damoyolo-t | 41.9% | 58.5% | 6.5 | 154.5ms | 8.5M | 18.1 | 2.3 | Model |
46 | damoyolo-nl | 40.5% | 57.9% | 19.4 | 51.6ms | 5.7M | 6.0 | 6.7 | Model |
47 | yolox-s | 40.4% | 59.4% | 43.5 | 23.0ms | 9.0M | 13.5 | 3.0 | Model |
48 | deimv2-pico | 38.5% | 55.3% | 51.6 | 19.4ms | 1.5M | 5.2 | 7.5 | Model |
49 | damoyolo-nm | 38.1% | 55.1% | 15.6 | 64.0ms | 2.7M | 3.7 | 10.3 | Model |
50 | yolox-tiny | 32.6% | 50.3% | 43.4 | 23.1ms | 5.1M | 7.7 | 4.3 | Model |
51 | damoyolo-ns | 32.2% | 48.0% | 17.8 | 56.2ms | 1.4M | 1.6 | 20.7 | Model |
52 | deimv2-femto | 31.0% | 46.0% | 57.2 | 17.5ms | 1.0M | 1.7 | 18.6 | Model |
53 | yolox-nano | 25.8% | 41.6% | 41.1 | 24.3ms | 0.9M | 1.3 | 19.5 | Model |
54 | deimv2-atto | 23.8% | 36.7% | 59.7 | 16.8ms | 0.5M | 0.8 | 31.3 | Model |
VA v1 Score
The composite ranking is coming back, but it will stay unpublished until the reviewed submission set is broad enough to make the ranking credible.
HardwareNVIDIA A100
RuntimePyTorch FP32
D-FINERF-DETRRT-DETRDEIMYOLOX
Preview only
25 of 25 modelsi
Vision AnalysisUltralytics(YOLO11, YOLOv8)
Roboflow(RF-DETR)
Tsinghua(YOLOv10)
Baidu(RT-DETR)
Academia Sinica(YOLOv9)
Megvii(YOLOX)
Coming soon
Composite ranking in progress
VA v1 Score Over Time
The historical timeline is returning as part of the same composite score rollout. The chart stays visible as a preview, but the live series is not published yet.
Ultralytics
Megvii
Academia Sinica
Tsinghua
Roboflow
Baidu
Open Source
Coming soon