From bfc8b72e4e2a1612b4aa4c12f7b3db9cdd7b34c4 Mon Sep 17 00:00:00 2001 From: Xingyi Zhou Date: Sun, 20 Nov 2022 15:07:51 -0800 Subject: [PATCH] update model zoo links --- docs/MODEL_ZOO.md | 30 +++++++++++++++--------------- 1 file changed, 15 insertions(+), 15 deletions(-) diff --git a/docs/MODEL_ZOO.md b/docs/MODEL_ZOO.md index 7a2a92b6..97063b95 100644 --- a/docs/MODEL_ZOO.md +++ b/docs/MODEL_ZOO.md @@ -5,7 +5,7 @@ - Multiscale training is used by default in all models. The results are all reported using single-scale testing. - We report runtime on our local workstation with a TitanXp GPU and a Titan RTX GPU. - All models are trained on 8-GPU servers by default. The 1280 models are trained on 24G GPUs. Reducing the batchsize with the linear learning rate rule should be fine. -- All models can be downloaded directly from [Google drive](https://drive.google.com/drive/folders/1eae1cTX8tvIaCeof36sBgxrXEXALYlf-?usp=sharing). +- All models can be downloaded directly from [Google drive](https://drive.google.com/drive/folders/1meZIsz8E3Ia9CRxLOAULDLeYrKMhhjJE). ## COCO @@ -14,8 +14,8 @@ | Model | val mAP | FPS (Titan Xp/ Titan RTX) | links | |-------------------------------------------|---------|---------|-----------| -| CenterNet-S4_DLA_8x | 42.5 | 50 / 71 |[config](../configs/CenterNet-S4_DLA_8x.yaml)/[model](https://drive.google.com/file/d/1lNBhVHnZAEBRD66MFaHjm5Ij6Z4KYrJq/view?usp=sharing)| -| CenterNet-FPN_R50_1x | 40.2 | 20 / 24 |[config](../configs/CenterNet-FPN_R50_1x.yaml)/[model](https://drive.google.com/file/d/1rVG1YTthMXvutC6jr9KoE2DthT5-jhGj/view?usp=sharing)| +| CenterNet-S4_DLA_8x | 42.5 | 50 / 71 |[config](../configs/CenterNet-S4_DLA_8x.yaml)/[model](https://drive.google.com/file/d/1AVfs9OoLePk_sqTPvqdRi1cXmO2cD0W_)| +| CenterNet-FPN_R50_1x | 40.2 | 20 / 24 |[config](../configs/CenterNet-FPN_R50_1x.yaml)/[model](https://drive.google.com/file/d/1iYlmjsBt9YIcaI8NzEwiMoaDDMHRmcR9)| #### Note @@ -33,14 +33,14 @@ | Model | val mAP | FPS (Titan Xp/ Titan RTX) | links | |-------------------------------------------|---------|---------|-----------| | CenterNet2-F_R50_1x | 41.7 | 22 / 27 |[config](../configs/CenterNet2-F_R50_1x.yaml)/[model](X)| -| CenterNet2_R50_1x | 42.9 | 18 / 24 |[config](../configs/CenterNet2_R50_1x.yaml)/[model](https://drive.google.com/file/d/1Osu1J_sskt_1FaGdfJKa4vd2N71TWS9W/view?usp=sharing)| -| CenterNet2_X101-DCN_2x | 49.9 | 6 / 8 |[config](../configs/CenterNet2_X101-DCN_2x.yaml)/[model](https://drive.google.com/file/d/1IHgpUHVJWpvMuFUUetgKWsw27pRNN2oK/view?usp=sharing)| -| CenterNet2_DLA-BiFPN-P3_4x | 43.8 | 40 / 50|[config](../configs/CenterNet2_DLA-BiFPN-P3_4x.yaml)/[model](https://drive.google.com/file/d/12GUNlDW9RmOs40UEMSiiUsk5QK_lpGsE/view?usp=sharing)| -| CenterNet2_DLA-BiFPN-P3_24x | 45.6 | 40 / 50 |[config](../configs/CenterNet2_DLA-BiFPN-P3_24x.yaml)/[model](https://drive.google.com/file/d/15ZES1ySxubDPzKsHPA7pYg8o_Vwmf-Mb/view?usp=sharing)| -| CenterNet2_R2-101-DCN_896_4x | 51.2 | 9 / 13 |[config](../configs/CenterNet2_R2-101-DCN_896_4x.yaml)/[model](https://drive.google.com/file/d/1S7_GE8ZDQBWuLEfKHkxzeF3KBsxsbABg/view?usp=sharing)| -| CenterNet2_R2-101-DCN-BiFPN_1280_4x | 52.9 | 6 / 8 |[config](../configs/CenterNet2_R2-101-DCN-BiFPN_1280_4x.yaml)/[model](https://drive.google.com/file/d/14EBHNMagBCNTQjOXcHoZwLYIi2lFIm7F/view?usp=sharing)| -| CenterNet2_R2-101-DCN-BiFPN_4x+4x_1560_ST | 56.1 | 3 / 5 |[config](../configs/CenterNet2_R2-101-DCN-BiFPN_4x+4x_1560_ST.yaml)/[model](https://drive.google.com/file/d/11ww9VlOi_nhpdsU_vBAecSxBU0dR_JzW/view?usp=sharing)| -| CenterNet2_DLA-BiFPN-P5_640_24x_ST | 49.2 | 33 / 38 |[config](../configs/CenterNet2_DLA-BiFPN-P5_640_24x_ST.yaml)/[model](https://drive.google.com/file/d/1qsHp2HrM1u8WrtBzF5S0oCoLMz-B40wk/view?usp=sharing)| +| CenterNet2_R50_1x | 42.9 | 18 / 24 |[config](../configs/CenterNet2_R50_1x.yaml)/[model](https://drive.google.com/file/d/1Qn0E_F1cmXtKPEdyZ_lSt-bnM9NueQpq)| +| CenterNet2_X101-DCN_2x | 49.9 | 6 / 8 |[config](../configs/CenterNet2_X101-DCN_2x.yaml)/[model](https://drive.google.com/file/d/1yuJbIlUgMiXdaDWRWArcsRsSoHti9e1y)| +| CenterNet2_DLA-BiFPN-P3_4x | 43.8 | 40 / 50|[config](../configs/CenterNet2_DLA-BiFPN-P3_4x.yaml)/[model](https://drive.google.com/file/d/1UGrnOE0W8Tgu6ffcCOQEbeUgThtDkbuQ)| +| CenterNet2_DLA-BiFPN-P3_24x | 45.6 | 40 / 50 |[config](../configs/CenterNet2_DLA-BiFPN-P3_24x.yaml)/[model](https://drive.google.com/file/d/17osgvr_Zhp9SS2uMa_YLiKwkKJIDtwPZ)| +| CenterNet2_R2-101-DCN_896_4x | 51.2 | 9 / 13 |[config](../configs/CenterNet2_R2-101-DCN_896_4x.yaml)/[model](https://drive.google.com/file/d/1YiJm7UtMstl63E8I4qQ8owteYC5zRFuQ)| +| CenterNet2_R2-101-DCN-BiFPN_1280_4x | 52.9 | 6 / 8 |[config](../configs/CenterNet2_R2-101-DCN-BiFPN_1280_4x.yaml)/[model](https://drive.google.com/file/d/1BIfEH04Lm3EvW9ov76yEPntUOJxaVoKd)| +| CenterNet2_R2-101-DCN-BiFPN_4x+4x_1560_ST | 56.1 | 3 / 5 |[config](../configs/CenterNet2_R2-101-DCN-BiFPN_4x+4x_1560_ST.yaml)/[model](https://drive.google.com/file/d/1GZyzJLB3FTcs8C7MpZRQWw44liYPyOMD)| +| CenterNet2_DLA-BiFPN-P5_640_24x_ST | 49.2 | 33 / 38 |[config](../configs/CenterNet2_DLA-BiFPN-P5_640_24x_ST.yaml)/[model](https://drive.google.com/file/d/1pGXpnHhvi66my_p5dASTnTjvaaj0FEvE)| #### Note @@ -49,15 +49,15 @@ - `CenterNet2_DLA-BiFPN-P3_24x` is trained by repeating the `4x` schedule (starting from learning rate 0.01) 6 times. - R2 means [Res2Net](https://github.com/Res2Net/Res2Net-detectron2) backbone. To train Res2Net models, you need to download the ImageNet pre-trained weight [here](https://github.com/Res2Net/Res2Net-detectron2) and place it in `output/r2_101.pkl`. - The last 4 models in the table are trained with the EfficientDet-style resize-and-crop augmentation, instead of the default random resizing short edge in detectron2. We found this trains faster (per-iteration) and gives better performance under a long schedule. -- `_ST` means using [self-training](https://arxiv.org/abs/2006.06882) using pseudo-labels produced by [Scaled-YOLOv4](https://github.com/WongKinYiu/ScaledYOLOv4) on COCO unlabeled images, with a hard score threshold 0.5. Our processed pseudo-labels can be downloaded [here](https://drive.google.com/file/d/1LMBjtHhLp6dYf6MjwEQmzCLWQLkmWPpw/view?usp=sharing). +- `_ST` means using [self-training](https://arxiv.org/abs/2006.06882) using pseudo-labels produced by [Scaled-YOLOv4](https://github.com/WongKinYiu/ScaledYOLOv4) on COCO unlabeled images, with a hard score threshold 0.5. Our processed pseudo-labels can be downloaded [here](https://drive.google.com/file/d/1R9tHlUaIrujmK6T08yJ0T77b2XzekisC). - `CenterNet2_R2-101-DCN-BiFPN_4x+4x_1560_ST` finetunes from `CenterNet2_R2-101-DCN-BiFPN_1280_4x` for an additional `4x` schedule with the self-training data. It is trained under `1280x1280` but tested under `1560x1560`. ## LVIS v1 | Model | val mAP box | links | |-------------------------------------------|--------------|-----------| -| LVIS_CenterNet2_R50_1x | 26.5 |[config](../configs/LVIS_CenterNet2_R50_1x.yaml)/[model](https://drive.google.com/file/d/1gT9e-tNw8uzEBaCadQuoOOP2TEYa4kKP/view?usp=sharing)| -| LVIS_CenterNet2_R50_Fed_1x | 28.3 |[config](../configs/LVIS_CenterNet2_R50_Fed_1x.yaml)/[model](https://drive.google.com/file/d/1a9UjheMCKax0qAKEwPVpq2ZHN6vpqJv8/view?usp=sharing)| +| LVIS_CenterNet2_R50_1x | 26.5 |[config](../configs/LVIS_CenterNet2_R50_1x.yaml)/[model](https://drive.google.com/file/d/1oOOKEDQIWW19AHhfnTb7HYZ3Z9gkZn_K)| +| LVIS_CenterNet2_R50_Fed_1x | 28.3 |[config](../configs/LVIS_CenterNet2_R50_Fed_1x.yaml)/[model](https://drive.google.com/file/d/1ETurGA7KIC5XMkMBI8MOIMDD_iJyMTif)| - The models are trained with repeat-factor sampling. - `LVIS_CenterNet2_R50_Fed_1x` is CenterNet2 with our federated loss. Check our Appendix D of our [paper](https://arxiv.org/abs/2103.07461) or our [technical report at LVIS challenge](https://www.lvisdataset.org/assets/challenge_reports/2020/CenterNet2.pdf) for references. @@ -66,7 +66,7 @@ | Model | val mAP| links | |-------------------------------------------|---------|-----------| -| O365_CenterNet2_R50_1x | 22.6 |[config](../configs/O365_CenterNet2_R50_1x.yaml)/[model](https://drive.google.com/file/d/18fG6xGchAlpNp5sx8RAtwadGkS-gdIBU/view?usp=sharing)| +| O365_CenterNet2_R50_1x | 22.6 |[config](../configs/O365_CenterNet2_R50_1x.yaml)/[model](https://drive.google.com/file/d/11d1Qx75otBAQQL2raxMTVJb17Qr56M3O)| #### Note - Objects365 dataset can be downloaded [here](https://www.objects365.org/overview.html).