Skip to content

P
PaddleDetection

PaddlePaddle / PaddleDetection
9 个月 前同步成功

通知 691
Star 11112
Fork 2696
P
PaddleDetection

前往新版Gitcode，体验更适合开发者的 AI 搜索 >>

未验证 提交 9cd784b0 编写于 作者: W wangguanzhong 提交者： GitHub
浏览文件
操作

update speed (#985) 

* update speed

* update model size of blazeface nas_v2

* remove speed of retinanet

* update doc
上级 276bb001
隐藏空白更改
内联 并排
Showing with 78 addition and 71 deletion
configs/anchor_free/README.md
浏览文件 @ 9cd784b0
......@@ -25,14 +25,12 @@
| 网络结构 | 骨干网络 | 图片个数/GPU | 预训练模型 | mAP | FPS | 模型下载 | 配置文件 |
|:------------:|:--------:|:----:|:-------:|:-------:|:---------:|:----------:|:----------:|
| CornerNet-Squeeze | Hourglass104 | 14 | 无 | 34.5 | 35.5 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/cornernet_squeeze_hg104.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/anchor_free/cornernet_squeeze_hg104.yml) |
| CornerNet-Squeeze | ResNet50-vd | 14 | [faster\_rcnn\_r50\_vd\_fpn\_2x](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r50_vd_fpn_2x.tar) | 32.7 | 42.45 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/cornernet_squeeze_r50_vd_fpn.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/anchor_free/cornernet_squeeze_r50_vd_fpn.yml) |
| CornerNet-Squeeze-dcn | ResNet50-vd | 14 | [faster\_rcnn\_dcn\_r50\_vd\_fpn\_2x](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_dcn_r50_vd_fpn_2x.tar) | 34.9 | 40.05 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/cornernet_squeeze_dcn_r50_vd_fpn.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/anchor_free/cornernet_squeeze_dcn_r50_vd_fpn.yml) |
| CornerNet-Squeeze-dcn-mixup-cosine* | ResNet50-vd | 14 | [faster\_rcnn\_dcn\_r50\_vd\_fpn\_2x](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_dcn_r50_vd_fpn_2x.tar) | 38.2 | 40.05 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/cornernet_squeeze_dcn_r50_vd_fpn_mixup_cosine.pdparams) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/anchor_free/cornernet_squeeze_dcn_r50_vd_fpn_mixup_cosine.yml) |
| FCOS | ResNet50 | 2 | [ResNet50\_cos\_pretrained](https://paddle-imagenet-models-name.bj.bcebos.com/ResNet50_cos_pretrained.tar) | 39.8 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/fcos_r50_fpn_1x.pdparams) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/anchor_free/fcos_r50_fpn_1x.yml) |
| FCOS+multiscale_train | ResNet50 | 2 | [ResNet50\_cos\_pretrained](https://paddle-imagenet-models-name.bj.bcebos.com/ResNet50_cos_pretrained.tar) | 42.0 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/fcos_r50_fpn_multiscale_2x.pdparams) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/anchor_free/fcos_r50_fpn_multiscale_2x.yml) |
| FCOS+DCN | ResNet50 | 2 | [ResNet50\_cos\_pretrained](https://paddle-imagenet-models-name.bj.bcebos.com/ResNet50_cos_pretrained.tar) | 44.4 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/fcos_dcn_r50_fpn_1x.pdparams) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/anchor_free/fcos_dcn_r50_fpn_1x.yml) |
| FCOS+DCN+cutmix | ResNet50 | 2 | [ResNet50\_cos\_pretrained](https://paddle-imagenet-models-name.bj.bcebos.com/ResNet50_cos_pretrained.tar) | 44.5 | - | [下载链接]
(https://paddlemodels.bj.bcebos.com/object_detection/fcos_dcn_r50_fpn_1x_cutmix.pdparams) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/anchor_free/fcos_dcn_r50_fpn_1x_cutmix.yml) |
| CornerNet-Squeeze | ResNet50-vd | 14 | [faster\_rcnn\_r50\_vd\_fpn\_2x](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r50_vd_fpn_2x.tar) | 32.7 | 47.01 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/cornernet_squeeze_r50_vd_fpn.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/anchor_free/cornernet_squeeze_r50_vd_fpn.yml) |
| CornerNet-Squeeze-dcn | ResNet50-vd | 14 | [faster\_rcnn\_dcn\_r50\_vd\_fpn\_2x](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_dcn_r50_vd_fpn_2x.tar) | 34.9 | 40.43 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/cornernet_squeeze_dcn_r50_vd_fpn.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/anchor_free/cornernet_squeeze_dcn_r50_vd_fpn.yml) |
| CornerNet-Squeeze-dcn-mixup-cosine* | ResNet50-vd | 14 | [faster\_rcnn\_dcn\_r50\_vd\_fpn\_2x](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_dcn_r50_vd_fpn_2x.tar) | 38.2 | 39.70 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/cornernet_squeeze_dcn_r50_vd_fpn_mixup_cosine.pdparams) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/anchor_free/cornernet_squeeze_dcn_r50_vd_fpn_mixup_cosine.yml) |
| FCOS | ResNet50 | 2 | [ResNet50\_cos\_pretrained](https://paddle-imagenet-models-name.bj.bcebos.com/ResNet50_cos_pretrained.tar) | 39.8 | 18.85 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/fcos_r50_fpn_1x.pdparams) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/anchor_free/fcos_r50_fpn_1x.yml) |
| FCOS+multiscale_train | ResNet50 | 2 | [ResNet50\_cos\_pretrained](https://paddle-imagenet-models-name.bj.bcebos.com/ResNet50_cos_pretrained.tar) | 42.0 | 19.05 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/fcos_r50_fpn_multiscale_2x.pdparams) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/anchor_free/fcos_r50_fpn_multiscale_2x.yml) |
| FCOS+DCN | ResNet50 | 2 | [ResNet50\_cos\_pretrained](https://paddle-imagenet-models-name.bj.bcebos.com/ResNet50_cos_pretrained.tar) | 44.4 | 13.66 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/fcos_dcn_r50_fpn_1x.pdparams) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/anchor_free/fcos_dcn_r50_fpn_1x.yml) |
**注意:**
......
docs/MODEL_ZOO.md
浏览文件 @ 9cd784b0
......@@ -46,7 +46,7 @@ The backbone models pretrained on ImageNet are available. All backbone models ar
| ResNet50-FPN | Mask | 1 | 1x | 15.184 | 37.9 | 34.2 | [model](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_r50_fpn_1x.tar) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/mask_rcnn_r50_fpn_1x.yml) |
| ResNet50-FPN | Mask | 1 | 2x | 15.881 | 38.7 | 34.7 | [model](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_r50_fpn_2x.tar) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/mask_rcnn_r50_fpn_2x.yml) |
| ResNet50-FPN | Cascade Faster | 2 | 1x | 17.507 | 40.9 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/cascade_rcnn_r50_fpn_1x.tar) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/cascade_rcnn_r50_fpn_1x.yml) |
| ResNet50-FPN | Cascade Mask | 1 | 1x | - | 41.3 | 35.5 | [model](https://paddlemodels.bj.bcebos.com/object_detection/cascade_mask_rcnn_r50_fpn_1x.tar) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/cascade_mask_rcnn_r50_fpn_1x.yml) |
| ResNet50-FPN | Cascade Mask | 1 | 1x | 12.43 | 41.3 | 35.5 | [model](https://paddlemodels.bj.bcebos.com/object_detection/cascade_mask_rcnn_r50_fpn_1x.tar) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/cascade_mask_rcnn_r50_fpn_1x.yml) |
| ResNet50-vd-FPN | Faster | 2 | 2x | 21.847 | 38.9 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r50_vd_fpn_2x.tar) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/faster_rcnn_r50_vd_fpn_2x.yml) |
| ResNet50-vd-FPN | Mask | 1 | 2x | 15.825 | 39.8 | 35.4 | [model](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_r50_vd_fpn_2x.tar) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/mask_rcnn_r50_vd_fpn_2x.yml) |
| CBResNet50-vd-FPN | Faster | 2 | 1x | - | 39.7 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_cbr50_vd_dual_fpn_1x.tar) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/faster_rcnn_cbr50_vd_dual_fpn_1x.yml) |
......@@ -83,8 +83,8 @@ The backbone models pretrained on ImageNet are available. All backbone models ar
| ResNet101-vd-FPN | Cascade Faster | c3-c5 | 2 | 1x | - | 46.4 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/cascade_rcnn_dcn_r101_vd_fpn_1x.tar) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/dcn/cascade_rcnn_dcn_r101_vd_fpn_1x.yml) |
| ResNeXt101-vd-FPN | Cascade Faster | c3-c5 | 2 | 1x | - | 47.3 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/cascade_rcnn_dcn_x101_vd_64x4d_fpn_1x.tar) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/dcn/cascade_rcnn_dcn_x101_vd_64x4d_fpn_1x.yml) |
| SENet154-vd-FPN | Cascade Mask | c3-c5 | 1 | 1.44x | - | 51.9 | 43.9 | [model](https://paddlemodels.bj.bcebos.com/object_detection/cascade_mask_rcnn_dcnv2_se154_vd_fpn_gn_s1x.tar) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/dcn/cascade_mask_rcnn_dcnv2_se154_vd_fpn_gn_s1x.yml) |
| ResNet200-vd-FPN-Nonlocal | CascadeClsAware Faster | c3-c5 | 1 | 2.5x | - | 51.7%(softnms) | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/cascade_rcnn_cls_aware_r200_vd_fpn_dcnv2_nonlocal_softnms.tar) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/dcn/cascade_rcnn_cls_aware_r200_vd_fpn_dcnv2_nonlocal_softnms.yml) |
| CBResNet200-vd-FPN-Nonlocal | Cascade Faster | c3-c5 | 1 | 2.5x | - | 53.3%(softnms) | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/cascade_rcnn_cbr200_vd_fpn_dcnv2_nonlocal_softnms.tar) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/dcn/cascade_rcnn_cbr200_vd_fpn_dcnv2_nonlocal_softnms.yml) |
| ResNet200-vd-FPN-Nonlocal | CascadeClsAware Faster | c3-c5 | 1 | 2.5x | 3.103 | 51.7%(softnms) | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/cascade_rcnn_cls_aware_r200_vd_fpn_dcnv2_nonlocal_softnms.tar) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/dcn/cascade_rcnn_cls_aware_r200_vd_fpn_dcnv2_nonlocal_softnms.yml) |
| CBResNet200-vd-FPN-Nonlocal | Cascade Faster | c3-c5 | 1 | 2.5x | 1.68 | 53.3%(softnms) | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/cascade_rcnn_cbr200_vd_fpn_dcnv2_nonlocal_softnms.tar) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/dcn/cascade_rcnn_cbr200_vd_fpn_dcnv2_nonlocal_softnms.yml) |
**Notes:**
......@@ -173,11 +173,11 @@ results of image size 608/416/320 above. Deformable conv is added on stage 5 of
### RetinaNet
| Backbone | Image/gpu | Lr schd | Box AP | Download | Configs |
| :---------------: | :-----: | :-----: | :----: | :-------: | :----: |
| ResNet50-FPN | 2 | 1x | 36.0 | [model](https://paddlemodels.bj.bcebos.com/object_detection/retinanet_r50_fpn_1x.tar) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/retinanet_r50_fpn_1x.yml) |
| ResNet101-FPN | 2 | 1x | 37.3 | [model](https://paddlemodels.bj.bcebos.com/object_detection/retinanet_r101_fpn_1x.tar) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/retinanet_r101_fpn_1x.yml) |
| ResNeXt101-vd-FPN | 1 | 1x | 40.5 | [model](https://paddlemodels.bj.bcebos.com/object_detection/retinanet_x101_vd_64x4d_fpn_1x.tar) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/retinanet_x101_vd_64x4d_fpn_1x.yml) |
| Backbone | Image/gpu | Lr schd | Inf time (fps) | Box AP | Download | Configs |
| :---------------: | :-----: | :-----: | :----: | :----: | :-------: | :----: |
| ResNet50-FPN | 2 | 1x | - | 36.0 | [model](https://paddlemodels.bj.bcebos.com/object_detection/retinanet_r50_fpn_1x.tar) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/retinanet_r50_fpn_1x.yml) |
| ResNet101-FPN | 2 | 1x | - | 37.3 | [model](https://paddlemodels.bj.bcebos.com/object_detection/retinanet_r101_fpn_1x.tar) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/retinanet_r101_fpn_1x.yml) |
| ResNeXt101-vd-FPN | 1 | 1x | - | 40.5 | [model](https://paddlemodels.bj.bcebos.com/object_detection/retinanet_x101_vd_64x4d_fpn_1x.tar) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/retinanet_x101_vd_64x4d_fpn_1x.yml) |
**Notes:** In RetinaNet, the base LR is changed to 0.01 for minibatch size 16.
......
docs/MODEL_ZOO_cn.md
浏览文件 @ 9cd784b0
......@@ -43,7 +43,7 @@ Paddle提供基于ImageNet的骨架网络预训练模型。所有预训练模型
| ResNet50-FPN | Mask | 1 | 1x | 15.184 | 37.9 | 34.2 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_r50_fpn_1x.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/mask_rcnn_r50_fpn_1x.yml) |
| ResNet50-FPN | Mask | 1 | 2x | 15.881 | 38.7 | 34.7 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_r50_fpn_2x.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/mask_rcnn_r50_fpn_2x.yml) |
| ResNet50-FPN | Cascade Faster | 2 | 1x | 17.507 | 40.9 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/cascade_rcnn_r50_fpn_1x.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/cascade_rcnn_r50_fpn_1x.yml) |
| ResNet50-FPN | Cascade Mask | 1 | 1x | - | 41.3 | 35.5 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/cascade_mask_rcnn_r50_fpn_1x.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/cascade_mask_rcnn_r50_fpn_1x.yml) |
| ResNet50-FPN | Cascade Mask | 1 | 1x | 12.43 | 41.3 | 35.5 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/cascade_mask_rcnn_r50_fpn_1x.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/cascade_mask_rcnn_r50_fpn_1x.yml) |
| ResNet50-vd-FPN | Faster | 2 | 2x | 21.847 | 38.9 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r50_vd_fpn_2x.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/faster_rcnn_r50_vd_fpn_2x.yml) |
| ResNet50-vd-FPN | Mask | 1 | 2x | 15.825 | 39.8 | 35.4 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_r50_vd_fpn_2x.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/mask_rcnn_r50_vd_fpn_2x.yml) |
| CBResNet50-vd-FPN | Faster | 2 | 1x | - | 39.7 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_cbr50_vd_dual_fpn_1x.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/faster_rcnn_cbr50_vd_dual_fpn_1x.yml) |
......@@ -80,8 +80,8 @@ Paddle提供基于ImageNet的骨架网络预训练模型。所有预训练模型
| ResNet101-vd-FPN | Cascade Faster | c3-c5 | 2 | 1x | - | 46.4 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/cascade_rcnn_dcn_r101_vd_fpn_1x.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/dcn/cascade_rcnn_dcn_r101_vd_fpn_1x.yml) |
| ResNeXt101-vd-FPN | Cascade Faster | c3-c5 | 2 | 1x | - | 47.3 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/cascade_rcnn_dcn_x101_vd_64x4d_fpn_1x.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/dcn/cascade_rcnn_dcn_x101_vd_64x4d_fpn_1x.yml) |
| SENet154-vd-FPN | Cascade Mask | c3-c5 | 1 | 1.44x | - | 51.9 | 43.9 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/cascade_mask_rcnn_dcnv2_se154_vd_fpn_gn_s1x.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/dcn/cascade_mask_rcnn_dcnv2_se154_vd_fpn_gn_s1x.yml) |
| ResNet200-vd-FPN-Nonlocal | CascadeClsAware Faster | c3-c5 | 1 | 2.5x | - | 51.7%(softnms) | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/cascade_rcnn_cls_aware_r200_vd_fpn_dcnv2_nonlocal_softnms.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/dcn/cascade_rcnn_cls_aware_r200_vd_fpn_dcnv2_nonlocal_softnms.yml) |
| CBResNet200-vd-FPN-Nonlocal | Cascade Faster | c3-c5 | 1 | 2.5x | - | 53.3%(softnms) | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/cascade_rcnn_cbr200_vd_fpn_dcnv2_nonlocal_softnms.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/dcn/cascade_rcnn_cbr200_vd_fpn_dcnv2_nonlocal_softnms.yml) |
| ResNet200-vd-FPN-Nonlocal | CascadeClsAware Faster | c3-c5 | 1 | 2.5x | 3.103 | 51.7%(softnms) | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/cascade_rcnn_cls_aware_r200_vd_fpn_dcnv2_nonlocal_softnms.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/dcn/cascade_rcnn_cls_aware_r200_vd_fpn_dcnv2_nonlocal_softnms.yml) |
| CBResNet200-vd-FPN-Nonlocal | Cascade Faster | c3-c5 | 1 | 2.5x | 1.68 | 53.3%(softnms) | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/cascade_rcnn_cbr200_vd_fpn_dcnv2_nonlocal_softnms.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/dcn/cascade_rcnn_cbr200_vd_fpn_dcnv2_nonlocal_softnms.yml) |
**注意事项:**
......@@ -165,11 +165,11 @@ Paddle提供基于ImageNet的骨架网络预训练模型。所有预训练模型
### RetinaNet
| 骨架网络 | 每张GPU图片个数 | 学习率策略 | Box AP | 下载 | 配置文件 |
| :---------------: | :-----: | :-----: | :----: | :-------: | :----: |
| ResNet50-FPN | 2 | 1x | 36.0 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/retinanet_r50_fpn_1x.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/retinanet_r50_fpn_1x.yml) |
| ResNet101-FPN | 2 | 1x | 37.3 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/retinanet_r101_fpn_1x.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/retinanet_r101_fpn_1x.yml) |
| ResNeXt101-vd-FPN | 1 | 1x | 40.5 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/retinanet_x101_vd_64x4d_fpn_1x.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/retinanet_x101_vd_64x4d_fpn_1x.yml) |
| 骨架网络 | 每张GPU图片个数 | 学习率策略 | 推理时间(fps) | Box AP | 下载 | 配置文件 |
| :---------------: | :-----: | :-----: | :----: | :----: | :-------: | :----: |
| ResNet50-FPN | 2 | 1x | - | 36.0 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/retinanet_r50_fpn_1x.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/retinanet_r50_fpn_1x.yml) |
| ResNet101-FPN | 2 | 1x | - | 37.3 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/retinanet_r101_fpn_1x.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/retinanet_r101_fpn_1x.yml) |
| ResNeXt101-vd-FPN | 1 | 1x | - | 40.5 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/retinanet_x101_vd_64x4d_fpn_1x.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/retinanet_x101_vd_64x4d_fpn_1x.yml) |
**注意事项:** RetinaNet系列模型中，在总batch size为16下情况下，初始学习率改为0.01。
......
docs/featured_model/FACE_DETECTION.md
浏览文件 @ 9cd784b0
......@@ -59,32 +59,35 @@ FaceDetection的目标是提供高效、高速的人脸检测解决方案，包
#### 推理时间和模型大小比较
| 网络结构 | 类型 | 输入尺寸 | P4(trt32) (ms) | CPU (ms) | 高通骁龙855(armv8) (ms) | 模型大小(MB) |
|:------------:|:--------:|:----:|:--------------:|:--------:|:-------------------------------------:|:---------------:|
| BlazeFace | 原始版本 | 128 | 1.387 | 23.461 | 6.036 | 0.777 |
| BlazeFace | Lite版本 | 128 | 1.323 | 12.802 | 6.193 | 0.68 |
| BlazeFace | NAS版本 | 128 | 1.03 | 6.714 | 2.7152 | 0.234 |
| FaceBoxes | 原始版本 | 128 | 3.144 | 14.972 | 19.2196 | 3.6 |
| FaceBoxes | Lite版本 | 128 | 2.295 | 11.276 | 8.5278 | 2 |
| BlazeFace | 原始版本 | 320 | 3.01 | 132.408 | 70.6916 | 0.777 |
| BlazeFace | Lite版本 | 320 | 2.535 | 69.964 | 69.9438 | 0.68 |
| BlazeFace | NAS版本 | 320 | 2.392 | 36.962 | 39.8086 | 0.234 |
| FaceBoxes | 原始版本 | 320 | 7.556 | 84.531 | 52.1022 | 3.6 |
| FaceBoxes | Lite版本 | 320 | 18.605 | 78.862 | 59.8996 | 2 |
| BlazeFace | 原始版本 | 640 | 8.885 | 519.364 | 149.896 | 0.777 |
| BlazeFace | Lite版本 | 640 | 6.988 | 284.13 | 149.902 | 0.68 |
| BlazeFace | NAS版本 | 640 | 7.448 | 142.91 | 69.8266 | 0.234 |
| FaceBoxes | 原始版本 | 640 | 78.201 | 394.043 | 169.877 | 3.6 |
| FaceBoxes | Lite版本 | 640 | 59.47 | 313.683 | 139.918 | 2 |
| 网络结构 | 类型 | 输入尺寸 | P4(trt32) (ms) | CPU (ms) | CPU (ms)(enable_mkldmm) | 高通骁龙855(armv8) (ms) | 模型大小(MB) |
|:------------:|:--------:|:----:|:--------------:|:--------:|:--------:|:-------------------------------------:|:---------------:|
| BlazeFace | 原始版本 | 128 | 1.387 | 23.461 | 4.92 | 6.036 | 0.777 |
| BlazeFace | Lite版本 | 128 | 1.323 | 12.802 | 7.16 | 6.193 | 0.68 |
| BlazeFace | NAS版本 | 128 | 1.03 | 6.714 | 3.641 | 2.7152 | 0.234 |
| BlazeFace | NAS_V2版本 | 128 | 0.909 | 9.58 | 7.903 | 3.499 | 0.383 |
| FaceBoxes | 原始版本 | 128 | 3.144 | 14.972 | 9,852 | 19.2196 | 3.6 |
| FaceBoxes | Lite版本 | 128 | 2.295 | 11.276 | 6.969 | 8.5278 | 2 |
| BlazeFace | 原始版本 | 320 | 3.01 | 132.408 | 20.762 | 70.6916 | 0.777 |
| BlazeFace | Lite版本 | 320 | 2.535 | 69.964 | 35.612 | 69.9438 | 0.68 |
| BlazeFace | NAS版本 | 320 | 2.392 | 36.962 | 14.443 | 39.8086 | 0.234 |
| BlazeFace | NAS_V2版本 | 320 | 1.487 | 52.038 | 38.693 | 56.137 | 0.383 |
| FaceBoxes | 原始版本 | 320 | 7.556 | 84.531 | 48.465 | 52.1022 | 3.6 |
| FaceBoxes | Lite版本 | 320 | 18.605 | 78.862 | 46.488 | 59.8996 | 2 |
| BlazeFace | 原始版本 | 640 | 8.885 | 519.364 | 78.825 | 149.896 | 0.777 |
| BlazeFace | Lite版本 | 640 | 6.988 | 284.13 | 131.385 | 149.902 | 0.68 |
| BlazeFace | NAS版本 | 640 | 7.448 | 142.91 | 56.725 | 69.8266 | 0.234 |
| BlazeFace | NAS_V2版本 | 640 | 4.201 | 197.695 | 153.626 | 88.278 | 0.383 |
| FaceBoxes | 原始版本 | 640 | 78.201 | 394.043 | 239.201 | 169.877 | 3.6 |
| FaceBoxes | Lite版本 | 640 | 59.47 | 313.683 | 168.73 | 139.918 | 2 |
**注意:**
- CPU: Intel(R) Xeon(R) CPU E5-2650 v4 @ 2.20GHz。
- P4(trt32)和CPU的推理时间测试基于PaddlePaddle-1.6.1版本。
- P4(trt32)和CPU的推理时间测试基于PaddlePaddle-1.8.0版本。
- ARM测试环境:
- 高通骁龙855(armv8)；
- 单线程；
- Paddle-Lite 2.0.0版本。
- Paddle-Lite develop版本。
## 快速开始
......
docs/featured_model/FACE_DETECTION_en.md
浏览文件 @ 9cd784b0
......@@ -66,32 +66,34 @@ For details can refer to [Evaluation](#Evaluate-on-the-FDDB).
#### Infer Time and Model Size comparison
| Architecture | Type | Size | P4(trt32) (ms) | CPU (ms) | Qualcomm SnapDragon 855(armv8) (ms) | Model size (MB) |
|:------------:|:--------:|:----:|:--------------:|:--------:|:-------------------------------------:|:---------------:|
| BlazeFace | Original | 128 | 1.387 | 23.461 | 6.036 | 0.777 |
| BlazeFace | Lite | 128 | 1.323 | 12.802 | 6.193 | 0.68 |
| BlazeFace | NAS | 128 | 1.03 | 6.714 | 2.7152 | 0.234 |
| FaceBoxes | Original | 128 | 3.144 | 14.972 | 19.2196 | 3.6 |
| FaceBoxes | Lite | 128 | 2.295 | 11.276 | 8.5278 | 2 |
| BlazeFace | Original | 320 | 3.01 | 132.408 | 70.6916 | 0.777 |
| BlazeFace | Lite | 320 | 2.535 | 69.964 | 69.9438 | 0.68 |
| BlazeFace | NAS | 320 | 2.392 | 36.962 | 39.8086 | 0.234 |
| FaceBoxes | Original | 320 | 7.556 | 84.531 | 52.1022 | 3.6 |
| FaceBoxes | Lite | 320 | 18.605 | 78.862 | 59.8996 | 2 |
| BlazeFace | Original | 640 | 8.885 | 519.364 | 149.896 | 0.777 |
| BlazeFace | Lite | 640 | 6.988 | 284.13 | 149.902 | 0.68 |
| BlazeFace | NAS | 640 | 7.448 | 142.91 | 69.8266 | 0.234 |
| FaceBoxes | Original | 640 | 78.201 | 394.043 | 169.877 | 3.6 |
| FaceBoxes | Lite | 640 | 59.47 | 313.683 | 139.918 | 2 |
| Architecture | Type | Size | P4(trt32) (ms) | CPU (ms) | CPU (ms)(enable_mkldmm) | Qualcomm SnapDragon 855(armv8) (ms) | Model size (MB) |
|:------------:|:--------:|:----:|:--------------:|:--------:|:--------:|:-------------------------------------:|:---------------:|
| BlazeFace | 原始版本 | 128 | 1.387 | 23.461 | 4.92 | 6.036 | 0.777 |
| BlazeFace | Lite版本 | 128 | 1.323 | 12.802 | 7.16 | 6.193 | 0.68 |
| BlazeFace | NAS版本 | 128 | 1.03 | 6.714 | 3.641 | 2.7152 | 0.234 |
| BlazeFace | NAS_V2版本 | 128 | 0.909 | 9.58 | 7.903 | 3.499 | 0.383 |
| FaceBoxes | 原始版本 | 128 | 3.144 | 14.972 | 9,852 | 19.2196 | 3.6 |
| FaceBoxes | Lite版本 | 128 | 2.295 | 11.276 | 6.969 | 8.5278 | 2 |
| BlazeFace | 原始版本 | 320 | 3.01 | 132.408 | 20.762 | 70.6916 | 0.777 |
| BlazeFace | Lite版本 | 320 | 2.535 | 69.964 | 35.612 | 69.9438 | 0.68 |
| BlazeFace | NAS版本 | 320 | 2.392 | 36.962 | 14.443 | 39.8086 | 0.234 |
| BlazeFace | NAS_V2版本 | 320 | 1.487 | 52.038 | 38.693 | 56.137 | 0.383 |
| FaceBoxes | 原始版本 | 320 | 7.556 | 84.531 | 48.465 | 52.1022 | 3.6 |
| FaceBoxes | Lite版本 | 320 | 18.605 | 78.862 | 46.488 | 59.8996 | 2 |
| BlazeFace | 原始版本 | 640 | 8.885 | 519.364 | 78.825 | 149.896 | 0.777 |
| BlazeFace | Lite版本 | 640 | 6.988 | 284.13 | 131.385 | 149.902 | 0.68 |
| BlazeFace | NAS版本 | 640 | 7.448 | 142.91 | 56.725 | 69.8266 | 0.234 |
| BlazeFace | NAS_V2版本 | 640 | 4.201 | 197.695 | 153.626 | 88.278 | 0.383 |
| FaceBoxes | 原始版本 | 640 | 78.201 | 394.043 | 239.201 | 169.877 | 3.6 |
| FaceBoxes | Lite版本 | 640 | 59.47 | 313.683 | 168.73 | 139.918 | 2 |
**NOTES:**
- CPU: Intel(R) Xeon(R) CPU E5-2650 v4 @ 2.20GHz.
- P4(trt32) and CPU tests based on PaddlePaddle, PaddlePaddle version is 1.6.1.
- P4(trt32) and CPU tests based on PaddlePaddle, PaddlePaddle version is 1.8.0.
- ARM test environment:
- Qualcomm SnapDragon 855(armv8);
- Single thread;
- Paddle-Lite version 2.0.0.
- Paddle-Lite version develop.
## Quick Start
......
docs/featured_model/YOLOv3_ENHANCEMENT.md
浏览文件 @ 9cd784b0
......@@ -41,11 +41,15 @@ python tools/train.py -c configs/dcn/yolov3_r50vd_dcn_db_iouloss_obj365_pretrain
### 模型效果
| 模型 | 预训练模型 | 验证集 mAP | P4预测速度 | 下载 | 配置文件 |
| :--------------------------------------: | :----------------------------------------------------------: | :--------: | :------------------------------------: | :----------------------------------------------------------: | :--------: |
| YOLOv3 DarkNet | [DarkNet pretrain](https://paddle-imagenet-models-name.bj.bcebos.com/DarkNet53_pretrained.tar) | 38.9 | 原生：88.3ms<br>tensorRT-FP32: 42.5ms | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_darknet.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/yolov3_darknet.yml) |
| YOLOv3 ResNet50_vd DCN | [ImageNet pretrain](https://paddle-imagenet-models-name.bj.bcebos.com/ResNet50_vd_pretrained.tar) | 39.1 | 原生：74.4ms<br>tensorRT-FP32: 35.2ms | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r50vd_dcn_imagenet.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/dcn/yolov3_r50vd_dcn.yml) |
| YOLOv3 ResNet50_vd DCN | [Object365 pretrain](https://paddlemodels.bj.bcebos.com/object_detection/ResNet50_vd_dcn_db_obj365_pretrained.tar) | 42.5 | 原生：74.4ms<br>tensorRT-FP32: 35.2ms | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r50vd_dcn_obj365_v2.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/dcn/yolov3_r50vd_dcn_obj365_pretrained_coco.yml) |
| YOLOv3 ResNet50_vd DCN DropBlock | [Object365 pretrain](https://paddlemodels.bj.bcebos.com/object_detection/ResNet50_vd_dcn_db_obj365_pretrained.tar) | 42.8 | 原生：74.4ms<br/>tensorRT-FP32: 35.2ms | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r50vd_dcn_obj365_dropblock.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/dcn/yolov3_r50vd_dcn_db_obj365_pretrained_coco.yml) |
| YOLOv3 ResNet50_vd DCN DropBlock IoULoss | [Object365 pretrain](https://paddlemodels.bj.bcebos.com/object_detection/ResNet50_vd_dcn_db_obj365_pretrained.tar) | 43.2 | 原生：74.4ms<br/>tensorRT-FP32: 35.2ms | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r50vd_dcn_obj365_dropblock_iouloss.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/dcn/yolov3_r50vd_dcn_db_iouloss_obj365_pretrained_coco.yml) |
| YOLOv3 ResNet50_vd DCN DropBlock IoU-Aware | [Object365 pretrain](https://paddlemodels.bj.bcebos.com/object_detection/ResNet50_vd_dcn_db_obj365_pretrained.tar) | 43.6 | 原生：74.4ms<br/>tensorRT-FP32: 35.2ms | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r50vd_dcn_db_iouaware_obj365_pretrained_coco.pdparams) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/dcn/yolov3_r50vd_dcn_db_iouaware_obj365_pretrained_coco.yml) |
| 模型 | 预训练模型 | 验证集 mAP | V100 python 预测速度（FPS）<sup>[1](#1)</sup> | V100 paddle预测库速度（ms/image）<sup>[2](#2)</sup> | P4 paddle预测库速度(ms/image) <sup>[2](#2)</sup> | 下载 | 配置文件 |
| :--------------------------------------: | :----------------------------------------------------------: | :--------: | :--------: | :------------------------------------: | :----------------------------------------------------------: | :--------: | :--------: |
| YOLOv3 DarkNet | [DarkNet pretrain](https://paddle-imagenet-models-name.bj.bcebos.com/DarkNet53_pretrained.tar) | 38.9 | 48.55 | 原生：19.63<br>tensorRT-FP32: 18.01<br>tensorRT-FP16: 11.47 | 原生：54.10<br>tensorRT-FP32: 47.44 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_darknet.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/yolov3_darknet.yml) |
| YOLOv3 ResNet50_vd DCN | [ImageNet pretrain](https://paddle-imagenet-models-name.bj.bcebos.com/ResNet50_vd_pretrained.tar) | 39.1 | 50.80 | 原生：17.04<br>tensorRT-FP32: 16.28<br>tensorRT-FP16: 11.16 | 原生：40.01<br>tensorRT-FP32: 36.66 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r50vd_dcn_imagenet.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/dcn/yolov3_r50vd_dcn.yml) |
| YOLOv3 ResNet50_vd DCN | [Object365 pretrain](https://paddlemodels.bj.bcebos.com/object_detection/ResNet50_vd_dcn_db_obj365_pretrained.tar) | 42.5 | 50.41 | 原生：16.76<br>tensorRT-FP32: 16.04<br>tensorRT-FP16: 10.70 | 原生：39.64<br>tensorRT-FP32: 35.93 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r50vd_dcn_obj365_v2.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/dcn/yolov3_r50vd_dcn_obj365_pretrained_coco.yml) |
| YOLOv3 ResNet50_vd DCN DropBlock | [Object365 pretrain](https://paddlemodels.bj.bcebos.com/object_detection/ResNet50_vd_dcn_db_obj365_pretrained.tar) | 42.8 | 49.97 | 原生：16.55<br>tensorRT-FP32: 16.07<br>tensorRT-FP16: 10.69 | 原生：39.72<br/>tensorRT-FP32: 35.98 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r50vd_dcn_obj365_dropblock.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/dcn/yolov3_r50vd_dcn_db_obj365_pretrained_coco.yml) |
| YOLOv3 ResNet50_vd DCN DropBlock IoULoss | [Object365 pretrain](https://paddlemodels.bj.bcebos.com/object_detection/ResNet50_vd_dcn_db_obj365_pretrained.tar) | 43.2 | 49.91 | 原生：16.46<br>tensorRT-FP32: 15.83<br>tensorRT-FP16: 10.80 | 原生：39.58<br/>tensorRT-FP32: 35.61 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r50vd_dcn_obj365_dropblock_iouloss.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/dcn/yolov3_r50vd_dcn_db_iouloss_obj365_pretrained_coco.yml) |
| YOLOv3 ResNet50_vd DCN DropBlock IoU-Aware | [Object365 pretrain](https://paddlemodels.bj.bcebos.com/object_detection/ResNet50_vd_dcn_db_obj365_pretrained.tar) | 43.6 | 48.19 | 原生：17.74<br>tensorRT-FP32: 16.73<br>tensorRT-FP16: 11.74 | 原生：41.39<br/>tensorRT-FP32: 37.75 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r50vd_dcn_db_iouaware_obj365_pretrained_coco.pdparams) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/dcn/yolov3_r50vd_dcn_db_iouaware_obj365_pretrained_coco.yml) |
<a name="1">[1]</a>V100 python 预测速度是在一张Tesla V100的GPU上通过```tools/eval.py```测试所有验证集得到，单位是fps(图片数/秒), cuDNN版本是7.5，包括数据加载、网络前向执行和后处理, batch size是1。
<a name="2">[2]</a>paddle预测库测试时，输入图片大小为640x640； 去掉前10轮warmup时间，测试100轮的平均时间; 开启了参数FLAGS_cudnn_exhaustive_search=True；使用代码deploy/python/infer.py测试
Markdown is supported
0% 或 .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑！
想要评论请 注册 或