参考：https://zhuanlan.zhihu.com/p/97862053
手机19fps
https://github.com/nightsnack/YOLObile
https://github.com/ZJULearning/ttfnet
基于mmcv，用到了deform_conv_cuda
| TTFNet-18 (1x) | 1.8 | 112.2 | 25.9 | Download |
| TTFNet-18 (2x) | 3.6 | 112.3 | 28.1 | Download |
| TTFNet-18 (no-pretrain 10x) | - | 121.0 | 30.3 | Download |
| TTFNet-18 (10x) | - | 113.6 | 31.8 | Download |
| TTFNet-34 (2x) | 4.1 | 86.6 | 31.3 | Download |
| TTFNet-34 (no-pretrain 10x) | - | 89.2 | 33.2 | Download |
| TTFNet-34 (10x) | - | 88.4 | 35.3 | Download |
| TTFNet-53 (1x) | 3.1 | 54.8 | 32.9 | Download |
| TTFNet-53 (2x) | 6.1 | 54.4 | 35.1 | Download |
| TTFNet-53 (no-pretrain 10x) | - | 57.2 | 36.2 | Download |
| TTFNet-53 (10x) | 30.6 | 57.0 | 39.3 | Download |
RDSNet：用于目标检测和实例分割的新的深度网络，没有发明backbone
https://mp.csdn.net/console/editor/html/105925868
torch版，170m
https://github.com/wangsr126/RDSNet
deform_conv_cuda
很多cuda和cu，需要编译
LRF-Net：高速学习丰富的特征以进行Single-Shot目标检测《Learning Rich Features at High-Speed for Single-Shot Object Detection》时间：20191214（ICCV 2019）作者团队：天津大学&IIAI链接：http://openaccess.thecvf.com/content_ICCV_2019/papers/Wang_Learning_Rich_Features_at_High-Speed_for_Single-Shot_Object_Detection_ICCV_2019_paper.pdf代码：https://github.com/vaesl/LRF-Net注：超越YOLOV3、RetinaNet等Single-stage网络，速度高达52FPS！代码刚开源！(基于PyTorch)
https://github.com/vaesl/LRF-Net
模型290-300m
腾讯推出超强少样本目标检测算法，公开千类少样本检测训练集FSOD
https://github.com/fanq15/Few-Shot-Object-Detection-Dataset
还未开源
1万+
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
