From d76a168a875e9cb7766f68e9dd32ece6fbca7758 Mon Sep 17 00:00:00 2001 From: qingqing01 Date: Thu, 5 Sep 2019 11:58:41 +0800 Subject: [PATCH] Add benchmark (fps) in MODEL_ZOO doc (#3263) * Add benchmark (fps) in MODEL_ZOO doc --- docs/MODEL_ZOO.md | 151 +++++++++++++++++++------------------- docs/MODEL_ZOO_cn.md | 149 ++++++++++++++++++------------------- ppdet/utils/eval_utils.py | 27 ++++++- tools/eval.py | 20 +---- tools/train.py | 20 ++--- 5 files changed, 189 insertions(+), 178 deletions(-) diff --git a/docs/MODEL_ZOO.md b/docs/MODEL_ZOO.md index 305f67c59..db71b2b22 100644 --- a/docs/MODEL_ZOO.md +++ b/docs/MODEL_ZOO.md @@ -2,9 +2,9 @@ ## Environment - Python 2.7.1 -- PaddlePaddle 1.5 +- PaddlePaddle >=1.5 - CUDA 9.0 -- CUDNN 7.4 +- cuDNN >=7.4 - NCCL 2.1.2 ## Common settings @@ -13,6 +13,8 @@ - Batch Normalization layers in backbones are replaced by Affine Channel layers. - Unless otherwise noted, all ResNet backbones adopt the [ResNet-B](https://arxiv.org/pdf/1812.01187) variant.. - For RCNN and RetinaNet models, only horizontal flipping data augmentation was used in the training phase and no augmentations were used in the testing phase. +- **Inf time (fps)**: the inference time is measured with fps (image/s) on a single GPU (Tesla V100) with cuDNN 7.5 by running 'tools/eval.py' on all validation set, which including data loadding, network forward and post processing. The batch size is 1. + ## Training Schedules @@ -30,50 +32,50 @@ The backbone models pretrained on ImageNet are available. All backbone models ar ### Faster & Mask R-CNN -| Backbone | Type | Image/gpu | Lr schd | Box AP | Mask AP | Download | -| :------------------- | :------------- | :-----: | :-----: | :----: | :-----: | :----------------------------------------------------------: | -| ResNet50 | Faster | 1 | 1x | 35.2 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r50_1x.tar) | -| ResNet50 | Faster | 1 | 2x | 37.1 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r50_2x.tar) | -| ResNet50 | Mask | 1 | 1x | 36.5 | 32.2 | [model](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_r50_1x.tar) | -| ResNet50 | Mask | 1 | 2x | 38.2 | 33.4 | [model](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_r50_2x.tar) | -| ResNet50-vd | Faster | 1 | 1x | 36.4 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r50_vd_1x.tar) | -| ResNet50-FPN | Faster | 2 | 1x | 37.2 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r50_fpn_1x.tar) | -| ResNet50-FPN | Faster | 2 | 2x | 37.7 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r50_fpn_2x.tar) | -| ResNet50-FPN | Mask | 1 | 1x | 37.9 | 34.2 | [model](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_r50_fpn_1x.tar) | -| ResNet50-FPN | Mask | 1 | 2x | 38.7 | 34.7 | [model](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_r50_fpn_2x.tar) | -| ResNet50-FPN | Cascade Faster | 2 | 1x | 40.9 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/cascade_rcnn_r50_fpn_1x.tar) | -| 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) | -| ResNet50-vd-FPN | Faster | 2 | 2x | 38.9 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r50_vd_fpn_2x.tar) | -| ResNet50-vd-FPN | Mask | 1 | 2x | 39.8 | 35.4 | [model](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_r50_vd_fpn_2x.tar) | -| ResNet101 | Faster | 1 | 1x | 38.3 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r101_1x.tar) | -| ResNet101-FPN | Faster | 1 | 1x | 38.7 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r101_fpn_1x.tar) | -| ResNet101-FPN | Faster | 1 | 2x | 39.1 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r101_fpn_2x.tar) | -| ResNet101-FPN | Mask | 1 | 1x | 39.5 | 35.2 | [model](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_r101_fpn_1x.tar) | -| ResNet101-vd-FPN | Faster | 1 | 1x | 40.5 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r101_vd_fpn_1x.tar) | -| ResNet101-vd-FPN | Faster | 1 | 2x | 40.8 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r101_vd_fpn_2x.tar) | -| ResNet101-vd-FPN | Mask | 1 | 1x | 41.4 | 36.8 | [model](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_r101_vd_fpn_1x.tar) | -| ResNeXt101-vd-FPN | Faster | 1 | 1x | 42.2 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_x101_vd_64x4d_fpn_1x.tar) | -| ResNeXt101-vd-FPN | Faster | 1 | 2x | 41.7 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_x101_vd_64x4d_fpn_2x.tar) | -| ResNeXt101-vd-FPN | Mask | 1 | 1x | 42.9 | 37.9 | [model](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_x101_vd_64x4d_fpn_1x.tar) | -| ResNeXt101-vd-FPN | Mask | 1 | 2x | 42.6 | 37.6 | [model](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_x101_vd_64x4d_fpn_2x.tar) | -| SENet154-vd-FPN | Faster | 1 | 1.44x | 42.9 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_se154_vd_fpn_s1x.tar) | -| SENet154-vd-FPN | Mask | 1 | 1.44x | 44.0 | 38.7 | [model](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_se154_vd_fpn_s1x.tar) | +| Backbone | Type | Image/gpu | Lr schd | Inf time (fps) | Box AP | Mask AP | Download | +| :---------------------- | :------------- | :-------: | :-----: | :------------: | :----: | :-----: | :----------------------------------------------------------: | +| ResNet50 | Faster | 1 | 1x | 12.747 | 35.2 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r50_1x.tar) | +| ResNet50 | Faster | 1 | 2x | 12.686 | 37.1 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r50_2x.tar) | +| ResNet50 | Mask | 1 | 1x | 11.615 | 36.5 | 32.2 | [model](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_r50_1x.tar) | +| ResNet50 | Mask | 1 | 2x | 11.494 | 38.2 | 33.4 | [model](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_r50_2x.tar) | +| ResNet50-vd | Faster | 1 | 1x | 12.575 | 36.4 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r50_vd_1x.tar) | +| ResNet50-FPN | Faster | 2 | 1x | 22.273 | 37.2 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r50_fpn_1x.tar) | +| ResNet50-FPN | Faster | 2 | 2x | 22.297 | 37.7 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r50_fpn_2x.tar) | +| 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) | +| 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) | +| ResNet50-FPN | Cascade Faster | 2 | 1x | 17.507 | 40.9 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/cascade_rcnn_r50_fpn_1x.tar) | +| 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) | +| 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) | +| 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) | +| ResNet101 | Faster | 1 | 1x | 9.316 | 38.3 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r101_1x.tar) | +| ResNet101-FPN | Faster | 1 | 1x | 17.297 | 38.7 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r101_fpn_1x.tar) | +| ResNet101-FPN | Faster | 1 | 2x | 17.246 | 39.1 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r101_fpn_2x.tar) | +| ResNet101-FPN | Mask | 1 | 1x | 12.983 | 39.5 | 35.2 | [model](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_r101_fpn_1x.tar) | +| ResNet101-vd-FPN | Faster | 1 | 1x | 17.011 | 40.5 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r101_vd_fpn_1x.tar) | +| ResNet101-vd-FPN | Faster | 1 | 2x | 16.934 | 40.8 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r101_vd_fpn_2x.tar) | +| ResNet101-vd-FPN | Mask | 1 | 1x | 13.105 | 41.4 | 36.8 | [model](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_r101_vd_fpn_1x.tar) | +| ResNeXt101-vd-64x4d-FPN | Faster | 1 | 1x | 8.815 | 42.2 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_x101_vd_64x4d_fpn_1x.tar) | +| ResNeXt101-vd-64x4d-FPN | Faster | 1 | 2x | 8.809 | 41.7 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_x101_vd_64x4d_fpn_2x.tar) | +| ResNeXt101-vd-64x4d-FPN | Mask | 1 | 1x | 7.689 | 42.9 | 37.9 | [model](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_x101_vd_64x4d_fpn_1x.tar) | +| ResNeXt101-vd-64x4d-FPN | Mask | 1 | 2x | 7.859 | 42.6 | 37.6 | [model](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_x101_vd_64x4d_fpn_2x.tar) | +| SENet154-vd-FPN | Faster | 1 | 1.44x | 3.408 | 42.9 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_se154_vd_fpn_s1x.tar) | +| SENet154-vd-FPN | Mask | 1 | 1.44x | 3.233 | 44.0 | 38.7 | [model](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_se154_vd_fpn_s1x.tar) | ### Deformable ConvNets v2 -| Backbone | Type | Conv | Image/gpu | Lr schd | Box AP | Mask AP | Download | -| :------------------- | :------------- | :-----: |:--------: | :-----: | :----: | :-----: | :----------------------------------------------------------: | -| ResNet50-FPN | Faster | c3-c5 | 2 | 1x | 41.0 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_dcn_r50_fpn_1x.tar) | -| ResNet50-vd-FPN | Faster | c3-c5 | 2 | 2x | 42.4 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_dcn_r50_vd_fpn_2x.tar) | -| ResNet101-vd-FPN | Faster | c3-c5 | 2 | 1x | 44.1 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_dcn_r101_vd_fpn_1x.tar) | -| ResNeXt101-vd-FPN | Faster | c3-c5 | 1 | 1x | 45.2 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_dcn_x101_vd_64x4d_fpn_1x.tar) | -| ResNet50-FPN | Mask | c3-c5 | 1 | 1x | 41.9 | 37.3 | [model](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_dcn_r50_fpn_1x.tar) | -| ResNet50-vd-FPN | Mask | c3-c5 | 1 | 2x | 42.9 | 38.0 | [model](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_dcn_r50_vd_fpn_2x.tar) | -| ResNet101-vd-FPN | Mask | c3-c5 | 1 | 1x | 44.6 | 39.2 | [model](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_dcn_r101_vd_fpn_1x.tar) | -| ResNeXt101-vd-FPN | Mask | c3-c5 | 1 | 1x | 46.2 | 40.4 | [model](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_dcn_x101_vd_64x4d_fpn_1x.tar) | -| ResNet50-FPN | Cascade Faster | c3-c5 | 2 | 1x | 44.2 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/cascade_rcnn_dcn_r50_fpn_1x.tar) | -| 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) | -| 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) | +| Backbone | Type | Conv | Image/gpu | Lr schd | Inf time (fps) | Box AP | Mask AP | Download | +| :---------------------- | :------------- | :---: | :-------: | :-----: | :------------: | :----: | :-----: | :----------------------------------------------------------: | +| ResNet50-FPN | Faster | c3-c5 | 2 | 1x | 19.978 | 41.0 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_dcn_r50_fpn_1x.tar) | +| ResNet50-vd-FPN | Faster | c3-c5 | 2 | 2x | 19.222 | 42.4 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_dcn_r50_vd_fpn_2x.tar) | +| ResNet101-vd-FPN | Faster | c3-c5 | 2 | 1x | 14.477 | 44.1 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_dcn_r101_vd_fpn_1x.tar) | +| ResNeXt101-vd-64x4d-FPN | Faster | c3-c5 | 1 | 1x | 7.209 | 45.2 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_dcn_x101_vd_64x4d_fpn_1x.tar) | +| ResNet50-FPN | Mask | c3-c5 | 1 | 1x | 14.53 | 41.9 | 37.3 | [model](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_dcn_r50_fpn_1x.tar) | +| ResNet50-vd-FPN | Mask | c3-c5 | 1 | 2x | 14.832 | 42.9 | 38.0 | [model](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_dcn_r50_vd_fpn_2x.tar) | +| ResNet101-vd-FPN | Mask | c3-c5 | 1 | 1x | 11.546 | 44.6 | 39.2 | [model](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_dcn_r101_vd_fpn_1x.tar) | +| ResNeXt101-vd-64x4d-FPN | Mask | c3-c5 | 1 | 1x | 6.45 | 46.2 | 40.4 | [model](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_dcn_x101_vd_64x4d_fpn_1x.tar) | +| ResNet50-FPN | Cascade Faster | c3-c5 | 2 | 1x | - | 44.2 | - | [model](https://paddlemodels.bj.bcebos.com/object_detection/cascade_rcnn_dcn_r50_fpn_1x.tar) | +| 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) | +| 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) | #### Notes: - Deformable ConvNets v2(dcn_v2) reference from [Deformable ConvNets v2](https://arxiv.org/abs/1811.11168). @@ -92,31 +94,32 @@ The backbone models pretrained on ImageNet are available. All backbone models ar ### Yolo v3 -| Backbone | Size | Image/gpu | Lr schd | Box AP | Download | -| :----------- | :--: | :-----: | :-----: | :----: | :-------: | -| DarkNet53 | 608 | 8 | 270e | 38.9 | [model](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_darknet.tar) | -| DarkNet53 | 416 | 8 | 270e | 37.5 | [model](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_darknet.tar) | -| DarkNet53 | 320 | 8 | 270e | 34.8 | [model](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_darknet.tar) | -| MobileNet-V1 | 608 | 8 | 270e | 29.3 | [model](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v1.tar) | -| MobileNet-V1 | 416 | 8 | 270e | 29.3 | [model](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v1.tar) | -| MobileNet-V1 | 320 | 8 | 270e | 27.1 | [model](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v1.tar) | -| ResNet34 | 608 | 8 | 270e | 36.2 | [model](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r34.tar) | -| ResNet34 | 416 | 8 | 270e | 34.3 | [model](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r34.tar) | -| ResNet34 | 320 | 8 | 270e | 31.4 | [model](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r34.tar) | +| Backbone | Size | Image/gpu | Lr schd | Inf time (fps) | Box AP | Download | +| :----------- | :--: | :-------: | :-----: | :------------: | :----: | :----------------------------------------------------------: | +| DarkNet53 | 608 | 8 | 270e | 45.571 | 38.9 | [model](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_darknet.tar) | +| DarkNet53 | 416 | 8 | 270e | - | 37.5 | [model](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_darknet.tar) | +| DarkNet53 | 320 | 8 | 270e | - | 34.8 | [model](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_darknet.tar) | +| MobileNet-V1 | 608 | 8 | 270e | 78.302 | 29.3 | [model](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v1.tar) | +| MobileNet-V1 | 416 | 8 | 270e | - | 29.3 | [model](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v1.tar) | +| MobileNet-V1 | 320 | 8 | 270e | - | 27.1 | [model](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v1.tar) | +| ResNet34 | 608 | 8 | 270e | 63.356 | 36.2 | [model](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r34.tar) | +| ResNet34 | 416 | 8 | 270e | - | 34.3 | [model](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r34.tar) | +| ResNet34 | 320 | 8 | 270e | - | 31.4 | [model](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r34.tar) | + ### Yolo v3 on Pascal VOC -| Backbone | Size | Image/gpu | Lr schd | Box AP | Download | -| :----------- | :--: | :-----: | :-----: | :----: | :-------: | -| DarkNet53 | 608 | 8 | 270e | 83.5 | [model](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_darknet_voc.tar) | -| DarkNet53 | 416 | 8 | 270e | 83.6 | [model](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_darknet_voc.tar) | -| DarkNet53 | 320 | 8 | 270e | 82.2 | [model](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_darknet_voc.tar) | -| MobileNet-V1 | 608 | 8 | 270e | 76.2 | [model](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v1_voc.tar) | -| MobileNet-V1 | 416 | 8 | 270e | 76.7 | [model](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v1_voc.tar) | -| MobileNet-V1 | 320 | 8 | 270e | 75.3 | [model](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v1_voc.tar) | -| ResNet34 | 608 | 8 | 270e | 82.6 | [model](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r34_voc.tar) | -| ResNet34 | 416 | 8 | 270e | 81.9 | [model](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r34_voc.tar) | -| ResNet34 | 320 | 8 | 270e | 80.1 | [model](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r34_voc.tar) | +| Backbone | Size | Image/gpu | Lr schd | Inf time (fps) | Box AP | Download | +| :----------- | :--: | :-------: | :-----: | :------------: | :----: | :----------------------------------------------------------: | +| DarkNet53 | 608 | 8 | 270e | 54.977 | 83.5 | [model](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_darknet_voc.tar) | +| DarkNet53 | 416 | 8 | 270e | - | 83.6 | [model](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_darknet_voc.tar) | +| DarkNet53 | 320 | 8 | 270e | - | 82.2 | [model](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_darknet_voc.tar) | +| MobileNet-V1 | 608 | 8 | 270e | 104.291 | 76.2 | [model](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v1_voc.tar) | +| MobileNet-V1 | 416 | 8 | 270e | - | 76.7 | [model](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v1_voc.tar) | +| MobileNet-V1 | 320 | 8 | 270e | - | 75.3 | [model](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v1_voc.tar) | +| ResNet34 | 608 | 8 | 270e | 82.247 | 82.6 | [model](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r34_voc.tar) | +| ResNet34 | 416 | 8 | 270e | - | 81.9 | [model](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r34_voc.tar) | +| ResNet34 | 320 | 8 | 270e | - | 80.1 | [model](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r34_voc.tar) | **Notes:** Yolo v3 is trained in 8 GPU with total batch size as 64 and trained 270 epoches. Yolo v3 training data augmentations: mixup, randomly color distortion, randomly cropping, randomly expansion, randomly interpolation method, randomly flippling. Yolo v3 used randomly @@ -135,20 +138,20 @@ results of image size 608/416/320 above. ### SSD -| Backbone | Size | Image/gpu | Lr schd | Box AP | Download | -| :----------: | :--: | :-------: | :-----: | :----: | :-------: | -| VGG16 | 300 | 8 | 40w | 25.1 | [model](https://paddlemodels.bj.bcebos.com/object_detection/ssd_vgg16_300.tar) | -| VGG16 | 512 | 8 | 40w | 29.1 | [model](https://paddlemodels.bj.bcebos.com/object_detection/ssd_vgg16_512.tar) | +| Backbone | Size | Image/gpu | Lr schd | Inf time (fps) | Box AP | Download | +| :------: | :--: | :-------: | :-----: | :------------: | :----: | :----------------------------------------------------------: | +| VGG16 | 300 | 8 | 40w | 81.613 | 25.1 | [model](https://paddlemodels.bj.bcebos.com/object_detection/ssd_vgg16_300.tar) | +| VGG16 | 512 | 8 | 40w | 46.007 | 29.1 | [model](https://paddlemodels.bj.bcebos.com/object_detection/ssd_vgg16_512.tar) | **Notes:** VGG-SSD is trained in 4 GPU with total batch size as 32 and trained 400000 iters. ### SSD on Pascal VOC -| Backbone | Size | Image/gpu | Lr schd | Box AP | Download | -| :----------- | :--: | :-----: | :-----: | :----: | :-------: | -| MobileNet v1 | 300 | 32 | 120e | 73.2 | [model](https://paddlemodels.bj.bcebos.com/object_detection/ssd_mobilenet_v1_voc.tar) | -| VGG16 | 300 | 8 | 240e | 77.5 | [model](https://paddlemodels.bj.bcebos.com/object_detection/ssd_vgg16_300_voc.tar) | -| VGG16 | 512 | 8 | 240e | 80.2 | [model](https://paddlemodels.bj.bcebos.com/object_detection/ssd_vgg16_512_voc.tar) | +| Backbone | Size | Image/gpu | Lr schd | Inf time (fps) | Box AP | Download | +| :----------- | :--: | :-------: | :-----: | :------------: | :----: | :----------------------------------------------------------: | +| MobileNet v1 | 300 | 32 | 120e | 159.543 | 73.2 | [model](https://paddlemodels.bj.bcebos.com/object_detection/ssd_mobilenet_v1_voc.tar) | +| VGG16 | 300 | 8 | 240e | 117.279 | 77.5 | [model](https://paddlemodels.bj.bcebos.com/object_detection/ssd_vgg16_300_voc.tar) | +| VGG16 | 512 | 8 | 240e | 65.975 | 80.2 | [model](https://paddlemodels.bj.bcebos.com/object_detection/ssd_vgg16_512_voc.tar) | **NOTE**: MobileNet-SSD is trained in 2 GPU with totoal batch size as 64 and trained 120 epoches. VGG-SSD is trained in 4 GPU with total batch size as 32 and trained 240 epoches. SSD training data augmentations: randomly color distortion, randomly cropping, randomly expansion, randomly flipping. diff --git a/docs/MODEL_ZOO_cn.md b/docs/MODEL_ZOO_cn.md index 57456b9f3..86450ad0a 100644 --- a/docs/MODEL_ZOO_cn.md +++ b/docs/MODEL_ZOO_cn.md @@ -3,9 +3,9 @@ ## 测试环境 - Python 2.7.1 -- PaddlePaddle 1.5 +- PaddlePaddle >=1.5 - CUDA 9.0 -- CUDNN 7.4 +- cuDNN >=7.4 - NCCL 2.1.2 ## 通用设置 @@ -13,6 +13,7 @@ - 所有模型均在COCO17数据集中训练和测试。 - 除非特殊说明，所有ResNet骨干网络采用[ResNet-B](https://arxiv.org/pdf/1812.01187)结构。 - 对于RCNN和RetinaNet系列模型，训练阶段仅使用水平翻转作为数据增强，测试阶段不使用数据增强。 +- **推理时间(fps)**: 推理时间是在一张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。 ## 训练策略 @@ -30,50 +31,50 @@ Paddle提供基于ImageNet的骨架网络预训练模型。所有预训练模型 ### Faster & Mask R-CNN -| 骨架网络 | 网络类型 | 每张GPU图片个数 | 学习率策略 | Box AP | Mask AP | 下载 | -| :------------------- | :------------- | :-----: | :-----: | :----: | :-----: | :----------------------------------------------------------: | -| ResNet50 | Faster | 1 | 1x | 35.2 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r50_1x.tar) | -| ResNet50 | Faster | 1 | 2x | 37.1 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r50_2x.tar) | -| ResNet50 | Mask | 1 | 1x | 36.5 | 32.2 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_r50_1x.tar) | -| ResNet50 | Mask | 1 | 2x | 38.2 | 33.4 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_r50_2x.tar) | -| ResNet50-vd | Faster | 1 | 1x | 36.4 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r50_vd_1x.tar) | -| ResNet50-FPN | Faster | 2 | 1x | 37.2 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r50_fpn_1x.tar) | -| ResNet50-FPN | Faster | 2 | 2x | 37.7 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r50_fpn_2x.tar) | -| ResNet50-FPN | Mask | 1 | 1x | 37.9 | 34.2 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_r50_fpn_1x.tar) | -| ResNet50-FPN | Mask | 1 | 2x | 38.7 | 34.7 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_r50_fpn_2x.tar) | -| ResNet50-FPN | Cascade Faster | 2 | 1x | 40.9 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/cascade_rcnn_r50_fpn_1x.tar) | -| 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) | -| ResNet50-vd-FPN | Faster | 2 | 2x | 38.9 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r50_vd_fpn_2x.tar) | -| ResNet50-vd-FPN | Mask | 1 | 2x | 39.8 | 35.4 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_r50_vd_fpn_2x.tar) | -| ResNet101 | Faster | 1 | 1x | 38.3 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r101_1x.tar) | -| ResNet101-FPN | Faster | 1 | 1x | 38.7 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r101_fpn_1x.tar) | -| ResNet101-FPN | Faster | 1 | 2x | 39.1 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r101_fpn_2x.tar) | -| ResNet101-FPN | Mask | 1 | 1x | 39.5 | 35.2 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_r101_fpn_1x.tar) | -| ResNet101-vd-FPN | Faster | 1 | 1x | 40.5 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r101_vd_fpn_1x.tar) | -| ResNet101-vd-FPN | Faster | 1 | 2x | 40.8 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r101_vd_fpn_2x.tar) | -| ResNet101-vd-FPN | Mask | 1 | 1x | 41.4 | 36.8 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_r101_vd_fpn_1x.tar) | -| ResNeXt101-vd-FPN | Faster | 1 | 1x | 42.2 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_x101_vd_64x4d_fpn_1x.tar) | -| ResNeXt101-vd-FPN | Faster | 1 | 2x | 41.7 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_x101_vd_64x4d_fpn_2x.tar) | -| ResNeXt101-vd-FPN | Mask | 1 | 1x | 42.9 | 37.9 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_x101_vd_64x4d_fpn_1x.tar) | -| ResNeXt101-vd-FPN | Mask | 1 | 2x | 42.6 | 37.6 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_x101_vd_64x4d_fpn_2x.tar) | -| SENet154-vd-FPN | Faster | 1 | 1.44x | 42.9 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_se154_vd_fpn_s1x.tar) | -| SENet154-vd-FPN | Mask | 1 | 1.44x | 44.0 | 38.7 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_se154_vd_fpn_s1x.tar) | +| 骨架网络 | 网络类型 | 每张GPU图片个数 | 学习率策略 |推理时间(fps) | Box AP | Mask AP | 下载 | +| :------------------- | :------------- | :-----: | :-----: | :------------: | :-----: | :-----: | :-----------------------------------------------------: | +| ResNet50 | Faster | 1 | 1x | 12.747 | 35.2 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r50_1x.tar) | +| ResNet50 | Faster | 1 | 2x | 12.686 | 37.1 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r50_2x.tar) | +| ResNet50 | Mask | 1 | 1x | 11.615 | 36.5 | 32.2 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_r50_1x.tar) | +| ResNet50 | Mask | 1 | 2x | 11.494 | 38.2 | 33.4 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_r50_2x.tar) | +| ResNet50-vd | Faster | 1 | 1x | 12.575 | 36.4 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r50_vd_1x.tar) | +| ResNet50-FPN | Faster | 2 | 1x | 22.273 | 37.2 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r50_fpn_1x.tar) | +| ResNet50-FPN | Faster | 2 | 2x | 22.297 | 37.7 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r50_fpn_2x.tar) | +| 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) | +| 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) | +| 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) | +| 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) | +| 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) | +| 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) | +| ResNet101 | Faster | 1 | 1x | 9.316 | 38.3 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r101_1x.tar) | +| ResNet101-FPN | Faster | 1 | 1x | 17.297 | 38.7 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r101_fpn_1x.tar) | +| ResNet101-FPN | Faster | 1 | 2x | 17.246 | 39.1 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r101_fpn_2x.tar) | +| ResNet101-FPN | Mask | 1 | 1x | 12.983 | 39.5 | 35.2 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_r101_fpn_1x.tar) | +| ResNet101-vd-FPN | Faster | 1 | 1x | 17.011 | 40.5 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r101_vd_fpn_1x.tar) | +| ResNet101-vd-FPN | Faster | 1 | 2x | 16.934 | 40.8 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_r101_vd_fpn_2x.tar) | +| ResNet101-vd-FPN | Mask | 1 | 1x | 13.105 | 41.4 | 36.8 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_r101_vd_fpn_1x.tar) | +| ResNeXt101-vd-FPN | Faster | 1 | 1x | 8.815 | 42.2 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_x101_vd_64x4d_fpn_1x.tar) | +| ResNeXt101-vd-FPN | Faster | 1 | 2x | 8.809 | 41.7 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_x101_vd_64x4d_fpn_2x.tar) | +| ResNeXt101-vd-FPN | Mask | 1 | 1x | 7.689 | 42.9 | 37.9 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_x101_vd_64x4d_fpn_1x.tar) | +| ResNeXt101-vd-FPN | Mask | 1 | 2x | 7.859 | 42.6 | 37.6 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_x101_vd_64x4d_fpn_2x.tar) | +| SENet154-vd-FPN | Faster | 1 | 1.44x | 3.408 | 42.9 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_se154_vd_fpn_s1x.tar) | +| SENet154-vd-FPN | Mask | 1 | 1.44x | 3.233 | 44.0 | 38.7 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_se154_vd_fpn_s1x.tar) | ### Deformable 卷积网络v2 -| 骨架网络 | 网络类型 | 卷积 | 每张GPU图片个数 | 学习率策略 | Box AP | Mask AP | 下载 | -| :------------------- | :------------- | :-----: |:--------: | :-----: | :----: | :-----: | :----------------------------------------------------------: | -| ResNet50-FPN | Faster | c3-c5 | 2 | 1x | 41.0 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_dcn_r50_fpn_1x.tar) | -| ResNet50-vd-FPN | Faster | c3-c5 | 2 | 2x | 42.4 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_dcn_r50_vd_fpn_2x.tar) | -| ResNet101-vd-FPN | Faster | c3-c5 | 2 | 1x | 44.1 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_dcn_r101_vd_fpn_1x.tar) | -| ResNeXt101-vd-FPN | Faster | c3-c5 | 1 | 1x | 45.2 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_dcn_x101_vd_64x4d_fpn_1x.tar) | -| ResNet50-FPN | Mask | c3-c5 | 1 | 1x | 41.9 | 37.3 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_dcn_r50_fpn_1x.tar) | -| ResNet50-vd-FPN | Mask | c3-c5 | 1 | 2x | 42.9 | 38.0 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_dcn_r50_vd_fpn_2x.tar) | -| ResNet101-vd-FPN | Mask | c3-c5 | 1 | 1x | 44.6 | 39.2 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_dcn_r101_vd_fpn_1x.tar) | -| ResNeXt101-vd-FPN | Mask | c3-c5 | 1 | 1x | 46.2 | 40.4 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_dcn_x101_vd_64x4d_fpn_1x.tar) | -| ResNet50-FPN | Cascade Faster | c3-c5 | 2 | 1x | 44.2 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/cascade_rcnn_dcn_r50_fpn_1x.tar) | -| 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) | -| 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) | +| 骨架网络 | 网络类型 | 卷积 | 每张GPU图片个数 | 学习率策略 |推理时间(fps)| Box AP | Mask AP | 下载 | +| :------------------- | :------------- | :-----: |:--------: | :-----: | :-----------: |:----: | :-----: | :----------------------------------------------------------: | +| ResNet50-FPN | Faster | c3-c5 | 2 | 1x | 19.978 | 41.0 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_dcn_r50_fpn_1x.tar) | +| ResNet50-vd-FPN | Faster | c3-c5 | 2 | 2x | 19.222 | 42.4 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_dcn_r50_vd_fpn_2x.tar) | +| ResNet101-vd-FPN | Faster | c3-c5 | 2 | 1x | 14.477 | 44.1 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_dcn_r101_vd_fpn_1x.tar) | +| ResNeXt101-vd-FPN | Faster | c3-c5 | 1 | 1x | 7.209 | 45.2 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/faster_rcnn_dcn_x101_vd_64x4d_fpn_1x.tar) | +| ResNet50-FPN | Mask | c3-c5 | 1 | 1x | 14.53 | 41.9 | 37.3 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_dcn_r50_fpn_1x.tar) | +| ResNet50-vd-FPN | Mask | c3-c5 | 1 | 2x | 14.832 | 42.9 | 38.0 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_dcn_r50_vd_fpn_2x.tar) | +| ResNet101-vd-FPN | Mask | c3-c5 | 1 | 1x | 11.546 | 44.6 | 39.2 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_dcn_r101_vd_fpn_1x.tar) | +| ResNeXt101-vd-FPN | Mask | c3-c5 | 1 | 1x | 6.45 | 46.2 | 40.4 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/mask_rcnn_dcn_x101_vd_64x4d_fpn_1x.tar) | +| ResNet50-FPN | Cascade Faster | c3-c5 | 2 | 1x | - | 44.2 | - | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/cascade_rcnn_dcn_r50_fpn_1x.tar) | +| 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) | +| 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) | #### 注意事项: - Deformable卷积网络v2(dcn_v2)参考自论文[Deformable ConvNets v2](https://arxiv.org/abs/1811.11168). @@ -92,31 +93,31 @@ Paddle提供基于ImageNet的骨架网络预训练模型。所有预训练模型 ### Yolo v3 -| 骨架网络 | 输入尺寸 | 每张GPU图片个数 | 学习率策略 | Box AP | 下载 | -| :----------- | :--: | :-----: | :-----: | :----: | :-------: | -| DarkNet53 | 608 | 8 | 270e | 38.9 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_darknet.tar) | -| DarkNet53 | 416 | 8 | 270e | 37.5 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_darknet.tar) | -| DarkNet53 | 320 | 8 | 270e | 34.8 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_darknet.tar) | -| MobileNet-V1 | 608 | 8 | 270e | 29.3 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v1.tar) | -| MobileNet-V1 | 416 | 8 | 270e | 29.3 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v1.tar) | -| MobileNet-V1 | 320 | 8 | 270e | 27.1 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v1.tar) | -| ResNet34 | 608 | 8 | 270e | 36.2 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r34.tar) | -| ResNet34 | 416 | 8 | 270e | 34.3 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r34.tar) | -| ResNet34 | 320 | 8 | 270e | 31.4 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r34.tar) | +| 骨架网络 | 输入尺寸 | 每张GPU图片个数 | 学习率策略 |推理时间(fps)| Box AP | 下载 | +| :----------- | :--: | :-----: | :-----: |:------------: |:----: | :-------: | +| DarkNet53 | 608 | 8 | 270e | 45.571 | 38.9 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_darknet.tar) | +| DarkNet53 | 416 | 8 | 270e | - | 37.5 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_darknet.tar) | +| DarkNet53 | 320 | 8 | 270e | - | 34.8 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_darknet.tar) | +| MobileNet-V1 | 608 | 8 | 270e | 78.302 | 29.3 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v1.tar) | +| MobileNet-V1 | 416 | 8 | 270e | - | 29.3 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v1.tar) | +| MobileNet-V1 | 320 | 8 | 270e | - | 27.1 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v1.tar) | +| ResNet34 | 608 | 8 | 270e | 63.356 | 36.2 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r34.tar) | +| ResNet34 | 416 | 8 | 270e | - | 34.3 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r34.tar) | +| ResNet34 | 320 | 8 | 270e | - | 31.4 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r34.tar) | ### Yolo v3 基于Pasacl VOC数据集 -| 骨架网络 | 输入尺寸 | 每张GPU图片个数 | 学习率策略 | Box AP | 下载 | -| :----------- | :--: | :-----: | :-----: | :----: | :-------: | -| DarkNet53 | 608 | 8 | 270e | 83.5 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_darknet_voc.tar) | -| DarkNet53 | 416 | 8 | 270e | 83.6 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_darknet_voc.tar) | -| DarkNet53 | 320 | 8 | 270e | 82.2 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_darknet_voc.tar) | -| MobileNet-V1 | 608 | 8 | 270e | 76.2 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v1_voc.tar) | -| MobileNet-V1 | 416 | 8 | 270e | 76.7 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v1_voc.tar) | -| MobileNet-V1 | 320 | 8 | 270e | 75.3 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v1_voc.tar) | -| ResNet34 | 608 | 8 | 270e | 82.6 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r34_voc.tar) | -| ResNet34 | 416 | 8 | 270e | 81.9 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r34_voc.tar) | -| ResNet34 | 320 | 8 | 270e | 80.1 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r34_voc.tar) | +| 骨架网络 | 输入尺寸 | 每张GPU图片个数 | 学习率策略 |推理时间(fps)| Box AP | 下载 | +| :----------- | :--: | :-----: | :-----: |:------------: |:----: | :-------: | +| DarkNet53 | 608 | 8 | 270e | 54.977 | 83.5 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_darknet_voc.tar) | +| DarkNet53 | 416 | 8 | 270e | - | 83.6 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_darknet_voc.tar) | +| DarkNet53 | 320 | 8 | 270e | - | 82.2 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_darknet_voc.tar) | +| MobileNet-V1 | 608 | 8 | 270e | 104.291 | 76.2 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v1_voc.tar) | +| MobileNet-V1 | 416 | 8 | 270e | - | 76.7 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v1_voc.tar) | +| MobileNet-V1 | 320 | 8 | 270e | - | 75.3 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v1_voc.tar) | +| ResNet34 | 608 | 8 | 270e | 82.247 | 82.6 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r34_voc.tar) | +| ResNet34 | 416 | 8 | 270e | - | 81.9 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r34_voc.tar) | +| ResNet34 | 320 | 8 | 270e | - | 80.1 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r34_voc.tar) | **注意事项:** Yolo v3在8卡，总batch size为64下训练270轮。数据增强包括：mixup, 随机颜色失真，随机剪裁，随机扩张，随机插值法，随机翻转。Yolo v3在训练阶段对minibatch采用随机reshape，可以采用相同的模型测试不同尺寸图片，我们分别提供了尺寸为608/416/320大小的测试结果。 @@ -132,19 +133,19 @@ Paddle提供基于ImageNet的骨架网络预训练模型。所有预训练模型 ### SSD -| 骨架网络 | 输入尺寸 | 每张GPU图片个数 | 学习率策略 | Box AP | 下载 | -| :----------: | :--: | :-------: | :-----: | :----: | :-------: | -| VGG16 | 300 | 8 | 40万 | 25.1 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/ssd_vgg16_300.tar) | -| VGG16 | 512 | 8 | 40万 | 29.1 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/ssd_vgg16_512.tar) | +| 骨架网络 | 输入尺寸 | 每张GPU图片个数 | 学习率策略|推理时间(fps) | Box AP | 下载 | +| :----------: | :--: | :-----: | :-----: |:------------: |:----: | :-------: | +| VGG16 | 300 | 8 | 40万 | 81.613 | 25.1 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/ssd_vgg16_300.tar) | +| VGG16 | 512 | 8 | 40万 | 46.007 | 29.1 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/ssd_vgg16_512.tar) | **注意事项:** VGG-SSD在总batch size为32下训练40万轮。 ### SSD 基于Pascal VOC数据集 -| 骨架网络 | 输入尺寸 | 每张GPU图片个数 | 学习率策略 | Box AP | 下载 | -| :----------- | :--: | :-----: | :-----: | :----: | :-------: | -| MobileNet v1 | 300 | 32 | 120e | 73.2 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/ssd_mobilenet_v1_voc.tar) | -| VGG16 | 300 | 8 | 240e | 77.5 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/ssd_vgg16_300_voc.tar) | -| VGG16 | 512 | 8 | 240e | 80.2 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/ssd_vgg16_512_voc.tar) | +| 骨架网络 | 输入尺寸 | 每张GPU图片个数 | 学习率策略 |推理时间(fps)| Box AP | 下载 | +| :----------- | :--: | :-----: | :-----: | :------------: |:----: | :-------: | +| MobileNet v1 | 300 | 32 | 120e | 159.543 | 73.2 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/ssd_mobilenet_v1_voc.tar) | +| VGG16 | 300 | 8 | 240e | 117.279 | 77.5 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/ssd_vgg16_300_voc.tar) | +| VGG16 | 512 | 8 | 240e | 65.975 | 80.2 | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/ssd_vgg16_512_voc.tar) | **注意事项:** MobileNet-SSD在2卡，总batch size为64下训练120周期。VGG-SSD在总batch size为32下训练240周期。数据增强包括：随机颜色失真，随机剪裁，随机扩张，随机翻转。 diff --git a/ppdet/utils/eval_utils.py b/ppdet/utils/eval_utils.py index afd3dbc7a..2af875e45 100644 --- a/ppdet/utils/eval_utils.py +++ b/ppdet/utils/eval_utils.py @@ -19,6 +19,7 @@ from __future__ import print_function import logging import numpy as np import os +import time import paddle.fluid as fluid @@ -69,6 +70,10 @@ def eval_run(exe, compile_program, pyreader, keys, values, cls): cls[i].reset(exe) values.append(accum_map) + images_num = 0 + start_time = time.time() + has_bbox = 'bbox' in keys + try: pyreader.start() while True: @@ -83,10 +88,20 @@ def eval_run(exe, compile_program, pyreader, keys, values, cls): if iter_id % 100 == 0: logger.info('Test iter {}'.format(iter_id)) iter_id += 1 + images_num += len(res['bbox'][1][0]) if has_bbox else 1 except (StopIteration, fluid.core.EOFException): pyreader.reset() logger.info('Test finish iter {}'.format(iter_id)) + end_time = time.time() + fps = images_num / (end_time - start_time) + if has_bbox: + logger.info('Total number of images: {}, inference time: {} fps.'. + format(images_num, fps)) + else: + logger.info('Total iteration: {}, inference time: {} batch/s.'.format( + images_num, fps)) + return results @@ -114,9 +129,12 @@ def eval_results(results, if output_directory: output = os.path.join(output_directory, 'bbox.json') - box_ap_stats = bbox_eval(results, anno_file, output, - with_background, - is_bbox_normalized=is_bbox_normalized) + box_ap_stats = bbox_eval( + results, + anno_file, + output, + with_background, + is_bbox_normalized=is_bbox_normalized) if 'mask' in results[0]: output = 'mask.json' @@ -130,7 +148,8 @@ def eval_results(results, box_ap_stats.append(res * 100.) elif 'bbox' in results[0]: box_ap = voc_bbox_eval( - results, num_classes, + results, + num_classes, is_bbox_normalized=is_bbox_normalized, map_type=map_type) box_ap_stats.append(box_ap) diff --git a/tools/eval.py b/tools/eval.py index d1be93908..eea606ec3 100644 --- a/tools/eval.py +++ b/tools/eval.py @@ -17,7 +17,6 @@ from __future__ import division from __future__ import print_function import os -import multiprocessing def set_paddle_flags(**kwargs): @@ -63,12 +62,6 @@ def main(): # check if set use_gpu=True in paddlepaddle cpu version check_gpu(cfg.use_gpu) - if cfg.use_gpu: - devices_num = fluid.core.get_cuda_device_count() - else: - devices_num = int( - os.environ.get('CPU_NUM', multiprocessing.cpu_count())) - if 'eval_feed' not in cfg: eval_feed = create(main_arch + 'EvalFeed') else: @@ -100,15 +93,9 @@ def main(): json_eval_results( eval_feed, cfg.metric, json_directory=FLAGS.output_eval) return - # compile program for multi-devices - if devices_num <= 1: - compile_program = fluid.compiler.CompiledProgram(eval_prog) - else: - build_strategy = fluid.BuildStrategy() - build_strategy.memory_optimize = False - build_strategy.enable_inplace = False - compile_program = fluid.compiler.CompiledProgram( - eval_prog).with_data_parallel(build_strategy=build_strategy) + + compile_program = fluid.compiler.CompiledProgram( + eval_prog).with_data_parallel() # load model exe.run(startup_prog) @@ -132,6 +119,7 @@ def main(): is_bbox_normalized = model.is_bbox_normalized() results = eval_run(exe, compile_program, pyreader, keys, values, cls) + # evaluation resolution = None if 'mask' in results[0]: diff --git a/tools/train.py b/tools/train.py index abcb7814a..95bba5fd2 100644 --- a/tools/train.py +++ b/tools/train.py @@ -18,16 +18,17 @@ from __future__ import print_function import os import time -import multiprocessing import numpy as np import datetime from collections import deque + def set_paddle_flags(**kwargs): for key, value in kwargs.items(): if os.environ.get(key, None) is None: os.environ[key] = str(value) + # NOTE(paddle-dev): All of these flags should be set before # `import paddle`. Otherwise, it would not take any effect. set_paddle_flags( @@ -69,8 +70,7 @@ def main(): if cfg.use_gpu: devices_num = fluid.core.get_cuda_device_count() else: - devices_num = int( - os.environ.get('CPU_NUM', multiprocessing.cpu_count())) + devices_num = int(os.environ.get('CPU_NUM', 1)) if 'train_feed' not in cfg: train_feed = create(main_arch + 'TrainFeed') @@ -133,12 +133,10 @@ def main(): # compile program for multi-devices build_strategy = fluid.BuildStrategy() - build_strategy.memory_optimize = False build_strategy.enable_inplace = False sync_bn = getattr(model.backbone, 'norm_type', None) == 'sync_bn' # only enable sync_bn in multi GPU devices - build_strategy.sync_batch_norm = sync_bn and devices_num > 1 \ - and cfg.use_gpu + build_strategy.sync_batch_norm = sync_bn and devices_num > 1 and cfg.use_gpu train_compile_program = fluid.compiler.CompiledProgram( train_prog).with_data_parallel( loss_name=loss.name, build_strategy=build_strategy) @@ -202,14 +200,16 @@ def main(): resolution = None if 'mask' in results[0]: resolution = model.mask_head.resolution - box_ap_stats = eval_results(results, eval_feed, cfg.metric, cfg.num_classes, - resolution, is_bbox_normalized, FLAGS.output_eval, map_type) + box_ap_stats = eval_results( + results, eval_feed, cfg.metric, cfg.num_classes, resolution, + is_bbox_normalized, FLAGS.output_eval, map_type) if box_ap_stats[0] > best_box_ap_list[0]: best_box_ap_list[0] = box_ap_stats[0] best_box_ap_list[1] = it - checkpoint.save(exe, train_prog, os.path.join(save_dir,"best_model")) + checkpoint.save(exe, train_prog, + os.path.join(save_dir, "best_model")) logger.info("Best test box ap: {}, in iter: {}".format( - best_box_ap_list[0],best_box_ap_list[1])) + best_box_ap_list[0], best_box_ap_list[1])) train_pyreader.reset() -- GitLab