From 5a50ccac6f5b10f1e5289475e00fbde5e146de62 Mon Sep 17 00:00:00 2001 From: lyuwenyu Date: Tue, 1 Aug 2023 11:46:48 +0800 Subject: [PATCH 1/4] add obj365 pretrained weights --- configs/rtdetr/README.md | 16 +++++++++++++++- 1 file changed, 15 insertions(+), 1 deletion(-) diff --git a/configs/rtdetr/README.md b/configs/rtdetr/README.md index cf5989ae93d..67baeae3111 100644 --- a/configs/rtdetr/README.md +++ b/configs/rtdetr/README.md @@ -20,7 +20,7 @@ RT-DETR是第一个实时端到端目标检测器。具体而言，我们设计 ## 基础模型 -| Model | Epoch | backbone | input shape | $AP^{val}$ | $AP^{val}_{50}$| Params(M) | FLOPs(G) | T4 TensorRT FP16(FPS) | Pretrained Model | config | +| Model | Epoch | Backbone | Input shape | $AP^{val}$ | $AP^{val}_{50}$| Params(M) | FLOPs(G) | T4 TensorRT FP16(FPS) | Pretrained Model | config | |:--------------:|:-----:|:----------:| :-------:|:--------------------------:|:---------------------------:|:---------:|:--------:| :---------------------: |:------------------------------------------------------------------------------------:|:-------------------------------------------:| | RT-DETR-R18 | 6x | ResNet-18 | 640 | 46.5 | 63.8 | 20 | 60 | 217 | [download](https://bj.bcebos.com/v1/paddledet/models/rtdetr_r18vd_dec3_6x_coco.pdparams) | [config](./rtdetr_r18vd_6x_coco.yml) | RT-DETR-R34 | 6x | ResNet-34 | 640 | 48.9 | 66.8 | 31 | 92 | 161 | [download](https://bj.bcebos.com/v1/paddledet/models/rtdetr_r34vd_dec4_6x_coco.pdparams) | [config](./rtdetr_r34vd_6x_coco.yml) @@ -37,6 +37,20 @@ RT-DETR是第一个实时端到端目标检测器。具体而言，我们设计 | RT-DETR-Swin | 3x | Swin_L_384 | 640 | 56.2 | 73.5 | [download](https://bj.bcebos.com/v1/paddledet/models/rtdetr_swin_L_384_3x_coco.pdparams) | [config](./rtdetr_swin_L_384_3x_coco.yml) | RT-DETR-FocalNet | 3x | FocalNet_L_384 | 640 | 56.9 | 74.3 | [download](https://bj.bcebos.com/v1/paddledet/models/rtdetr_focalnet_L_384_3x_coco.pdparams) | [config](./rtdetr_focalnet_L_384_3x_coco.yml) + +# Objects365预训练模型 + +## Model Zoo on Objects365 +| Model | Epoch | Dataset | Input shape | $AP^{val}$ | $AP^{val}_{50}$ | Weight +|:---:|:---:|:---:| :---:|:---:|:---:|:---:| +RT-DETR-R50 | 1x | Objects365 | 640 | 35.1 | 46.2 | [download](https://bj.bcebos.com/v1/paddledet/models/rtdetr_r50vd_1x_objects365.pdparams) +RT-DETR-R50 | 2x | COCO + Objects365 | 640 | 55.3 | 73.4 | [download](https://bj.bcebos.com/v1/paddledet/models/rtdetr_r50vd_2x_coco_objects365.pdparams) + +**Notes:** +- `COCO + Objects365` 代表使用Objects365预训练权重，在COCO上finetune的结果 + + + **注意事项:** - RT-DETR 基础模型均使用4个GPU训练。 - RT-DETR 在COCO train2017上训练，并在val2017上评估。 From 223e510fd365c811b6bcca744934afc0487b18d9 Mon Sep 17 00:00:00 2001 From: lyuwenyu Date: Tue, 1 Aug 2023 11:49:01 +0800 Subject: [PATCH 2/4] add obj365 pretrained weights --- configs/rtdetr/README.md | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/configs/rtdetr/README.md b/configs/rtdetr/README.md index 67baeae3111..d24e42589cc 100644 --- a/configs/rtdetr/README.md +++ b/configs/rtdetr/README.md @@ -41,10 +41,10 @@ RT-DETR是第一个实时端到端目标检测器。具体而言，我们设计 # Objects365预训练模型 ## Model Zoo on Objects365 -| Model | Epoch | Dataset | Input shape | $AP^{val}$ | $AP^{val}_{50}$ | Weight -|:---:|:---:|:---:| :---:|:---:|:---:|:---:| -RT-DETR-R50 | 1x | Objects365 | 640 | 35.1 | 46.2 | [download](https://bj.bcebos.com/v1/paddledet/models/rtdetr_r50vd_1x_objects365.pdparams) -RT-DETR-R50 | 2x | COCO + Objects365 | 640 | 55.3 | 73.4 | [download](https://bj.bcebos.com/v1/paddledet/models/rtdetr_r50vd_2x_coco_objects365.pdparams) +| Model | Epoch | Dataset | Input shape | $AP^{val}$ | $AP^{val}_{50}$ | Weight | Logs +|:---:|:---:|:---:| :---:|:---:|:---:|:---:|:---:| +RT-DETR-R50 | 1x | Objects365 | 640 | 35.1 | 46.2 | [download](https://bj.bcebos.com/v1/paddledet/models/rtdetr_r50vd_1x_objects365.pdparams) | [log](https://github.com/lyuwenyu/RT-DETR/issues/8) +RT-DETR-R50 | 2x | COCO + Objects365 | 640 | 55.3 | 73.4 | [download](https://bj.bcebos.com/v1/paddledet/models/rtdetr_r50vd_2x_coco_objects365.pdparams) | [log](https://github.com/lyuwenyu/RT-DETR/issues/8) **Notes:** - `COCO + Objects365` 代表使用Objects365预训练权重，在COCO上finetune的结果 From 2769ee836fdeec743335ca58aad53e8545f8049a Mon Sep 17 00:00:00 2001 From: lyuwenyu Date: Tue, 1 Aug 2023 11:56:29 +0800 Subject: [PATCH 3/4] add obj365 pretrained weights --- configs/rtdetr/README.md | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) diff --git a/configs/rtdetr/README.md b/configs/rtdetr/README.md index d24e42589cc..2cdd3db716b 100644 --- a/configs/rtdetr/README.md +++ b/configs/rtdetr/README.md @@ -38,9 +38,7 @@ RT-DETR是第一个实时端到端目标检测器。具体而言，我们设计 | RT-DETR-FocalNet | 3x | FocalNet_L_384 | 640 | 56.9 | 74.3 | [download](https://bj.bcebos.com/v1/paddledet/models/rtdetr_focalnet_L_384_3x_coco.pdparams) | [config](./rtdetr_focalnet_L_384_3x_coco.yml) -# Objects365预训练模型 - -## Model Zoo on Objects365 +## Objects365预训练模型 | Model | Epoch | Dataset | Input shape | $AP^{val}$ | $AP^{val}_{50}$ | Weight | Logs |:---:|:---:|:---:| :---:|:---:|:---:|:---:|:---:| RT-DETR-R50 | 1x | Objects365 | 640 | 35.1 | 46.2 | [download](https://bj.bcebos.com/v1/paddledet/models/rtdetr_r50vd_1x_objects365.pdparams) | [log](https://github.com/lyuwenyu/RT-DETR/issues/8) From 80ff53d50c9defbc88fc7023bc8557a25acf0875 Mon Sep 17 00:00:00 2001 From: lyuwenyu Date: Tue, 1 Aug 2023 12:18:35 +0800 Subject: [PATCH 4/4] add obj365 pretrained weights, test=document_fix --- configs/rtdetr/README.md | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/configs/rtdetr/README.md b/configs/rtdetr/README.md index 2cdd3db716b..7efdbd3c9ca 100644 --- a/configs/rtdetr/README.md +++ b/configs/rtdetr/README.md @@ -39,10 +39,10 @@ RT-DETR是第一个实时端到端目标检测器。具体而言，我们设计 ## Objects365预训练模型 -| Model | Epoch | Dataset | Input shape | $AP^{val}$ | $AP^{val}_{50}$ | Weight | Logs -|:---:|:---:|:---:| :---:|:---:|:---:|:---:|:---:| -RT-DETR-R50 | 1x | Objects365 | 640 | 35.1 | 46.2 | [download](https://bj.bcebos.com/v1/paddledet/models/rtdetr_r50vd_1x_objects365.pdparams) | [log](https://github.com/lyuwenyu/RT-DETR/issues/8) -RT-DETR-R50 | 2x | COCO + Objects365 | 640 | 55.3 | 73.4 | [download](https://bj.bcebos.com/v1/paddledet/models/rtdetr_r50vd_2x_coco_objects365.pdparams) | [log](https://github.com/lyuwenyu/RT-DETR/issues/8) +| Model | Epoch | Dataset | Input shape | $AP^{val}$ | $AP^{val}_{50}$ | T4 TensorRT FP16(FPS) | Weight | Logs +|:---:|:---:|:---:| :---:|:---:|:---:|:---:|:---:|:---:| +RT-DETR-R50 | 1x | Objects365 | 640 | 35.1 | 46.2 | 108 | [download](https://bj.bcebos.com/v1/paddledet/models/rtdetr_r50vd_1x_objects365.pdparams) | [log](https://github.com/lyuwenyu/RT-DETR/issues/8) +RT-DETR-R50 | 2x | COCO + Objects365 | 640 | 55.3 | 73.4 | 108 | [download](https://bj.bcebos.com/v1/paddledet/models/rtdetr_r50vd_2x_coco_objects365.pdparams) | [log](https://github.com/lyuwenyu/RT-DETR/issues/8) **Notes:** - `COCO + Objects365` 代表使用Objects365预训练权重，在COCO上finetune的结果