Table 3 Comparison with different models on MS COCO dataset.
From: LKC-Net: large kernel convolution object detection network
Models | AP@.5:0.95 (%) | Params (M) |
|---|---|---|
MNetV1-SSDLite37 | 22.2 | 5.10 |
MNetV2-SSDLite37 | 22.1 | 1.30 |
RefineDet512-VGG-1633 | 33.0 | – |
RFBNet512-VGG33 | 33.8 | – |
MnasNet-A1-SSDLite38 | 23 | 4.90 |
RetinaNet640-ResNet-5039 | 37.0 | – |
YOLOV3-ASFF32040 | 38.1 | – |
PPYOLO-Tiny41641 | 22.7 | 4.20 |
YOLOV4-Tiny32042 | 28.7 | 5.89 |
YOLOX-Tiny43 | 32.8 | 5.1 |
YOLOV5s29 | 37.2 | – |
YOLOV7-Tiny64036 | 37.4 | 6.2 |
DAMO-YOLO-Ns44 | 32.3 | 1.41 |
DAMO-YOLO-Nm44 | 38.2 | 2.14 |
PP-Picodet-M45 | 34.3 | 2.15 |
PP-PicoDet-MV3-large-1×45 | 35.6 | 2.80 |
PP-PicoDet-LCNet-1.5×45 | 36.3 | 3.10 |
EffificientDet-D0 (512)46 | 34.6 | 3.9 |
LKC-Net (Ours) | 38.4 | 7.2 |
