@@ -67,7 +67,7 @@ $ python track.py --source 0 --yolo-model checkpoints/yolov5s.pt --reid-model CL
67
67
68
68
Available ReID models (Feature Extractors):
69
69
* ** CLIP** : ` CLIP-RN50 ` , ` CLIP-ViT-B/32 `
70
- * ** DINO** : ` DINO-XciT-S12/16 ` , ` DINO-XciT-S12/8 ` , ` DINO-XciT- M24/16` , ` DINO-ViT-S/16 ` , ` DINO-ViT-S/8 ` , ` DINO-ViT-B/16 `
70
+ * ** DINO** : ` DINO-XciT-S12/16 ` , ` DINO-XciT-M24/16 ` , ` DINO-ViT-S/16 ` , ` DINO-ViT-B/16 `
71
71
72
72
Check [ here] ( tracking/utils.py#L14 ) to get COCO class index for your class.
73
73
@@ -110,20 +110,21 @@ YOLOv5m<sup><br>(CrowdHuman) | CLIP<sup><br>(RN50) | 53.25 | 43.25 | 52.12 | 912
110
110
YOLOv5m<sup ><br >(CrowdHuman) | CLIP<sup ><br >(ViT-B/32) | 53.35 | 43.03 | 51.25 | 896 | ** 199** | 91 | 14035 | ** 36575** | 4
111
111
||
112
112
YOLOv5m<sup ><br >(CrowdHuman) | DINO<sup ><br >(XciT-S12/16) | 54.41 | 47.44 | 59.01 | 511 | 184 | 101 | 12265 | 37555 |8
113
- YOLOv5m<sup ><br >(CrowdHuman) | DINO<sup ><br >(XciT-S12/8) | 54.44 | 47.63 | 59.24 | 517 | 185 | 98 | 12140 | 37639 | 4
114
- YOLOv5m<sup ><br >(CrowdHuman) | DINO<sup ><br >(XciT-M24/16) | 54.56 | ** 47.71** | ** 59.77** | 504 | 187 | 96 | 12364 | 37306 | 5
115
113
YOLOv5m<sup ><br >(CrowdHuman) | DINO<sup ><br >(ViT-S/16) | 54.56 | 47.61 | 58.94 | 519 | 189 | 97 | 12346 | 37308 | 8
116
- YOLOv5m<sup ><br >(CrowdHuman) | DINO<sup ><br >(ViT-S/8 ) | 54.53 | 47.70 | 59.20 | 542 | 180 | 102 | 11912 | 37744 | 4
114
+ YOLOv5m<sup ><br >(CrowdHuman) | DINO<sup ><br >(XciT-M24/16 ) | 54.56 | ** 47.71 ** | ** 59.77 ** | 504 | 187 | 96 | 12364 | 37306 | 5
117
115
YOLOv5m<sup ><br >(CrowdHuman) | DINO<sup ><br >(ViT-B/16) | ** 54.58** | 47.55 | 58.89 | 507 | 184 | 97 | 12017 | 37621 | 5
118
116
119
117
** FPS Results**
120
118
121
119
Detector | Feature Extractor | GPU | Precision | Image Size | Detection<br >/Frame | FPS
122
120
--- | --- | --- | --- | --- | --- | ---
123
- YOLOv5s | CLIP (RN50) | GTX-1660ti | FP32 | 480x640 | 1 | 40
124
- YOLOv5m | CLIP (RN50) | GTX-1660ti | FP32 | 480x640 | 1 | 32
125
- YOLOv5s | CLIP (ViT-B/32) | GTX-1660ti | FP32 | 480x640 | 1 | 30
126
- YOLOv5m | CLIP (ViT-B/32) | GTX-1660ti | FP32 | 480x640 | 1 | 23
121
+ YOLOv5s | CLIP-RN50 | GTX-1660ti | FP32 | 480x640 | 1 | 40
122
+ YOLOv5m | CLIP-RN50 | GTX-1660ti | FP32 | 480x640 | 1 | 32
123
+ YOLOv5s | CLIP-ViT-B/32 | GTX-1660ti | FP32 | 480x640 | 1 | 30
124
+ ||
125
+ YOLOv5s | DINO-XciT-S12/16 | GTX-1660ti | FP32 | 480x640 | 1 | 36
126
+ YOLOv5s | DINO-ViT-B/16 | GTX-1660ti | FP32 | 480x640 | 1 | 30
127
+ YOLOv5s | DINO-XciT-M24/16 | GTX-1660ti | FP32 | 480x640 | 1 | 25
127
128
128
129
129
130
## References
0 commit comments