Hugging Face's logo

dronefreak
/
visdrone-rfdetr-medium

Object Detection
PyTorch
rfdetr
detectionbench
computer-vision
aerial-imagery
drone
visdrone
detr
roboflow
transformer
Model card Files
xet
 Community
2

Upload README.md

#2
by dronefreak - opened
base: refs/heads/main
←
from: refs/pr/2
Discussion Files changed
+33
-33
Files changed (1) hide show
  1. README.md +33 -33
README.md CHANGED
@@ -91,45 +91,45 @@ Metrics reported in this model card are computed on the VisDrone-DET **test** sp
91
 
92
  Every model DetectionBench has trained and evaluated on VisDrone-DET so far, for full transparency -- see [DetectionBench](https://github.com/dronefreak/DetectionBench) for the smaller, curated comparison set used on the project README.
93
 
94
- | Rank | Model | mAP@50 | mAP@50-95 | Precision | Recall |
95
- | -------------------------- | --------------------- | ------------- | --------------- | ----------------- | -------------- |
96
- | 1 | RF-DETR Medium | 36.82 | 20.14 | 64.0 | 47.05 |
97
- | 2 | RF-DETR Small | 33.25 | 17.88 | 62.62 | 43.51 |
98
- | 3 | RF-DETR Nano | 25.15 | 12.77 | 58.99 | 35.0 |
99
  ---
100
 
101
  ## External VisDrone-DET Comparison
102
 
103
  The YOLO/RT-DETR rows below were trained and evaluated on VisDrone2019-DET's test split via a separate companion codebase ([VisDrone-dataset-python-toolkit](https://github.com/dronefreak/VisDrone-dataset-python-toolkit)), not reproduced inside DetectionBench -- included here purely for context. The RF-DETR rows are this repository's own DetectionBench-trained runs (see the Model Zoo table above).
104
 
105
- | Rank | Model | mAP@50 | mAP@50-95 | Precision | Recall |
106
- | ---- | ----- | ------ | --------- | --------- | ------ |
107
- | 1 | YOLOv9e | 40.02 | 23.73 | 54.78 | 42.42 |
108
- | 2 | YOLOv11x | 38.44 | 22.6 | 52.41 | 41.43 |
109
- | 3 | YOLOv26x | 38.33 | 22.48 | 52.91 | 41.06 |
110
- | 4 | YOLOv11l | 37.14 | 21.85 | 51.87 | 40.33 |
111
- | 5 | YOLOv10x | 37.24 | 21.81 | 52.59 | 39.84 |
112
- | 6 | YOLOv26l | 37.65 | 21.75 | 51.6 | 40.42 |
113
- | 7 | YOLOv9c | 37.22 | 21.73 | 51.99 | 39.77 |
114
- | 8 | YOLOv8x | 36.81 | 21.52 | 51.91 | 39.78 |
115
- | 9 | YOLOv26m | 36.67 | 21.22 | 51.03 | 39.79 |
116
- | 10 | YOLOv10l | 35.95 | 21.09 | 52.13 | 38.48 |
117
- | 11 | YOLOv11m | 36.35 | 21.02 | 50.24 | 39.46 |
118
- | 12 | YOLOv9m | 36.19 | 20.95 | 51.05 | 39.12 |
119
- | 13 | RF-DETR-Medium | 36.82 | 20.14 | 64.0 | 47.05 |
120
- | 14 | YOLOv8m | 34.39 | 19.95 | 48.18 | 38.2 |
121
- | 15 | YOLOv9s | 33.52 | 19.26 | 46.16 | 37.43 |
122
- | 16 | YOLOv11s | 32.3 | 18.47 | 45.49 | 35.31 |
123
- | 17 | YOLOv8s | 31.95 | 18.24 | 45.99 | 35.49 |
124
- | 18 | YOLOv26s | 32.1 | 18.06 | 45.75 | 35.05 |
125
- | 19 | RF-DETR-Small | 33.25 | 17.88 | 62.62 | 43.51 |
126
- | 20 | YOLOv9t | 29.09 | 16.22 | 42.57 | 32.66 |
127
- | 21 | YOLOv8n | 28.18 | 15.77 | 40.86 | 31.81 |
128
- | 22 | YOLOv11n | 27.59 | 15.46 | 39.58 | 31.74 |
129
- | 23 | YOLOv10n | 27.65 | 15.32 | 41.02 | 31.68 |
130
- | 24 | YOLOv26n | 26.73 | 14.64 | 38.6 | 31.14 |
131
- | 25 | RF-DETR-Nano | 25.15 | 12.77 | 58.99 | 35.0 |
132
- | 26 | rt_detr_l | 21.68 | 9.34 | 35.76 | 26.3 |
133
  Source: https://huggingface.co/collections/dronefreak/visdrone-detection-model-zoo
134
 
135
  ---
 
91
 
92
  Every model DetectionBench has trained and evaluated on VisDrone-DET so far, for full transparency -- see [DetectionBench](https://github.com/dronefreak/DetectionBench) for the smaller, curated comparison set used on the project README.
93
 
94
+ | Model | mAP@50 | mAP@50-95 | Precision | Recall |
95
+ | --------------------- | ------------- | --------------- | ----------------- | -------------- |
96
+ | RF-DETR Medium | 36.82 | 20.14 | 64.0 | 47.05 |
97
+ | RF-DETR Small | 33.25 | 17.88 | 62.62 | 43.51 |
98
+ | RF-DETR Nano | 25.15 | 12.77 | 58.99 | 35.0 |
99
  ---
100
 
101
  ## External VisDrone-DET Comparison
102
 
103
  The YOLO/RT-DETR rows below were trained and evaluated on VisDrone2019-DET's test split via a separate companion codebase ([VisDrone-dataset-python-toolkit](https://github.com/dronefreak/VisDrone-dataset-python-toolkit)), not reproduced inside DetectionBench -- included here purely for context. The RF-DETR rows are this repository's own DetectionBench-trained runs (see the Model Zoo table above).
104
 
105
+ | Model | mAP@50 | mAP@50-95 | Precision | Recall |
106
+ | ----- | ------ | --------- | --------- | ------ |
107
+ | YOLOv9e | 40.02 | 23.73 | 54.78 | 42.42 |
108
+ | YOLOv11x | 38.44 | 22.6 | 52.41 | 41.43 |
109
+ | YOLOv26x | 38.33 | 22.48 | 52.91 | 41.06 |
110
+ | YOLOv11l | 37.14 | 21.85 | 51.87 | 40.33 |
111
+ | YOLOv10x | 37.24 | 21.81 | 52.59 | 39.84 |
112
+ | YOLOv26l | 37.65 | 21.75 | 51.6 | 40.42 |
113
+ | YOLOv9c | 37.22 | 21.73 | 51.99 | 39.77 |
114
+ | YOLOv8x | 36.81 | 21.52 | 51.91 | 39.78 |
115
+ | YOLOv26m | 36.67 | 21.22 | 51.03 | 39.79 |
116
+ | YOLOv10l | 35.95 | 21.09 | 52.13 | 38.48 |
117
+ | YOLOv11m | 36.35 | 21.02 | 50.24 | 39.46 |
118
+ | YOLOv9m | 36.19 | 20.95 | 51.05 | 39.12 |
119
+ | RF-DETR-Medium | 36.82 | 20.14 | 64.0 | 47.05 |
120
+ | YOLOv8m | 34.39 | 19.95 | 48.18 | 38.2 |
121
+ | YOLOv9s | 33.52 | 19.26 | 46.16 | 37.43 |
122
+ | YOLOv11s | 32.3 | 18.47 | 45.49 | 35.31 |
123
+ | YOLOv8s | 31.95 | 18.24 | 45.99 | 35.49 |
124
+ | YOLOv26s | 32.1 | 18.06 | 45.75 | 35.05 |
125
+ | RF-DETR-Small | 33.25 | 17.88 | 62.62 | 43.51 |
126
+ | YOLOv9t | 29.09 | 16.22 | 42.57 | 32.66 |
127
+ | YOLOv8n | 28.18 | 15.77 | 40.86 | 31.81 |
128
+ | YOLOv11n | 27.59 | 15.46 | 39.58 | 31.74 |
129
+ | YOLOv10n | 27.65 | 15.32 | 41.02 | 31.68 |
130
+ | YOLOv26n | 26.73 | 14.64 | 38.6 | 31.14 |
131
+ | RF-DETR-Nano | 25.15 | 12.77 | 58.99 | 35.0 |
132
+ | rt_detr_l | 21.68 | 9.34 | 35.76 | 26.3 |
133
  Source: https://huggingface.co/collections/dronefreak/visdrone-detection-model-zoo
134
 
135
  ---