From ba9646fde34a201c72e5d4f530b7bd5dd7f7fb9c Mon Sep 17 00:00:00 2001 From: Yang Zhang Date: Thu, 14 May 2020 12:12:29 +0800 Subject: [PATCH] Add cn doc for mobile (#664) --- configs/mobile/README.md | 60 ++++++++++++++------------- configs/mobile/README_en.md | 82 +++++++++++++++++++++++++++++++++++++ 2 files changed, 113 insertions(+), 29 deletions(-) create mode 100755 configs/mobile/README_en.md diff --git a/configs/mobile/README.md b/configs/mobile/README.md index ff967105b..6cd8b8b65 100755 --- a/configs/mobile/README.md +++ b/configs/mobile/README.md @@ -1,39 +1,41 @@ -# Mobile Model Zoo +[English](README_en.md) | 简体中文 +# 移动端模型库 -## Models -This directory contains models optimized for mobile applications, at present the following models included: +## 模型 -| Backbone | Architecture | Input | Image/gpu 1 | Lr schd | Box AP | Download 2 | +PaddleDetection目前提供一系列针对移动应用进行优化的模型，主要支持以下结构: + +| 骨干网络 | 结构 | 输入大小 | 图片/gpu 1 | 学习率策略 | Box AP | 下载 2 | |--------------------------|---------------------------|-------|------------------------|---------------|--------|-----------------------| -| MobileNetV3 Small | SSDLite | 320 | 64 | 400K (cosine) | 16.6 | [Link](https://paddlemodels.bj.bcebos.com/object_detection/ssdlite_mobilenet_v3_small.tar.gz) | -| MobileNetV3 Large | SSDLite | 320 | 64 | 400K (cosine) | 22.8 | [Link](https://paddlemodels.bj.bcebos.com/object_detection/ssdlite_mobilenet_v3_large.tar.gz) | -| MobileNetV3 Large w/ FPN | Cascade RCNN | 320 | 2 | 500k (cosine) | 25.0 | [Link](https://paddlemodels.bj.bcebos.com/object_detection/cascade_rcnn_mobilenetv3_fpn_320.tar.gz) | -| MobileNetV3 Large w/ FPN | Cascade RCNN | 640 | 2 | 500k (cosine) | 30.2 | [Link](https://paddlemodels.bj.bcebos.com/object_detection/cascade_rcnn_mobilenetv3_fpn_640.tar.gz) | -| MobileNetV3 Large | YOLOv3 | 320 | 8 | 500K | 27.1 | [Link](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v3.tar.gz) | -| MobileNetV3 Large | YOLOv3 Prune 3 | 320 | 8 | - | 24.6 | [Link](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v3_prune86_FPGM_320.tar.gz) | +| MobileNetV3 Small | SSDLite | 320 | 64 | 400K (cosine) | 16.6 | [链接](https://paddlemodels.bj.bcebos.com/object_detection/ssdlite_mobilenet_v3_small.tar.gz) | +| MobileNetV3 Large | SSDLite | 320 | 64 | 400K (cosine) | 22.8 | [链接](https://paddlemodels.bj.bcebos.com/object_detection/ssdlite_mobilenet_v3_large.tar.gz) | +| MobileNetV3 Large w/ FPN | Cascade RCNN | 320 | 2 | 500k (cosine) | 25.0 | [链接](https://paddlemodels.bj.bcebos.com/object_detection/cascade_rcnn_mobilenetv3_fpn_320.tar.gz) | +| MobileNetV3 Large w/ FPN | Cascade RCNN | 640 | 2 | 500k (cosine) | 30.2 | [链接](https://paddlemodels.bj.bcebos.com/object_detection/cascade_rcnn_mobilenetv3_fpn_640.tar.gz) | +| MobileNetV3 Large | YOLOv3 | 320 | 8 | 500K | 27.1 | [链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v3.tar.gz) | +| MobileNetV3 Large | YOLOv3 Prune 3 | 320 | 8 | - | 24.6 | [链接](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v3_prune86_FPGM_320.tar.gz) | -**Notes**: +**注意**: -- [1] All models are trained on 8 GPUs. -- [2] Each tarball contains the following files - - model weight file (`.pdparams` or `.tar`) - - inference model files (`__model__` and `__params__`) - - Paddle-Lite model file (`.nb`) -- [3] See the note section on how YOLO head is pruned +- [1] 模型统一使用8卡训练. +- [2] 压缩包包括下列文件 + - 模型权重文件 (`.pdparams` or `.tar`) + - inference model 文件 (`__model__` and `__params__`) + - Paddle-Lite 模型文件 (`.nb`) +- [3] 参考下面关于YOLO剪裁的说明 -## Benchmarks Results +## 评测结果 -- Models are benched on following chipsets with [Paddle-Lite](https://github.com/PaddlePaddle/Paddle-Lite) 2.6 (to be released) +- 模型使用 [Paddle-Lite](https://github.com/PaddlePaddle/Paddle-Lite) 2.6 (即将发布) 在下列平台上进行了测试 - Qualcomm Snapdragon 625 - Qualcomm Snapdragon 835 - Qualcomm Snapdragon 845 - Qualcomm Snapdragon 855 - HiSilicon Kirin 970 - HiSilicon Kirin 980 -- With 1 CPU thread (latency numbers are in ms) +- 单CPU线程 (单位： ms) | | SD625 | SD835 | SD845 | SD855 | Kirin 970 | Kirin 980 | |------------------|---------|---------|---------|---------|-----------|-----------| @@ -43,7 +45,7 @@ This directory contains models optimized for mobile applications, at present the | YOLOv3 prune | 253.98 | 131.279 | 89.4124 | 48.2856 | 122.732 | 55.8626 | | Cascade RCNN 320 | 286.526 | 125.635 | 87.404 | 46.184 | 149.179 | 52.9994 | | Cascade RCNN 640 | 1115.66 | 495.926 | 351.361 | 189.722 | 573.558 | 207.917 | -- With 4 CPU threads (latency numbers are in ms) +- 4 CPU线程 (单位： ms) | | SD625 | SD835 | SD845 | SD855 | Kirin 970 | Kirin 980 | |------------------|---------|---------|---------|---------|-----------|-----------| @@ -55,26 +57,26 @@ This directory contains models optimized for mobile applications, at present the | Cascade RCNN 640 | 473.083 | 224.543 | 156.205 | 100.686 | 231.108 | 138.391 | -## Notes on YOLOv3 pruning +## YOLOv3剪裁说明 -We pruned the YOLO-head and distill the pruned model with YOLOv3-ResNet34 as the teacher, which has a higher mAP on COCO (31.4 with 320\*320 input). +首先对YOLO检测头进行剪裁，然后再使用 YOLOv3-ResNet34 作为teacher网络对剪裁后的模型进行蒸馏, teacher网络在COCO的mAP为31.4 (输入大小320\*320). -The following configurations can be used for pruning: +可以使用如下两种方式进行剪裁: -- Prune with fixed ratio, overall prune ratios is 86% +- 固定比例剪裁, 整体剪裁率是86% ```shell --pruned_params="yolo_block.0.0.0.conv.weights,yolo_block.0.0.1.conv.weights,yolo_block.0.1.0.conv.weights,yolo_block.0.1.1.conv.weights,yolo_block.0.2.conv.weights,yolo_block.0.tip.conv.weights,yolo_block.1.0.0.conv.weights,yolo_block.1.0.1.conv.weights,yolo_block.1.1.0.conv.weights,yolo_block.1.1.1.conv.weights,yolo_block.1.2.conv.weights,yolo_block.1.tip.conv.weights,yolo_block.2.0.0.conv.weights,yolo_block.2.0.1.conv.weights,yolo_block.2.1.0.conv.weights,yolo_block.2.1.1.conv.weights,yolo_block.2.2.conv.weights,yolo_block.2.tip.conv.weights" \ --pruned_ratios="0.75,0.75,0.75,0.75,0.75,0.75,0.75,0.75,0.75,0.75,0.75,0.75,0.875,0.875,0.875,0.875,0.875,0.875" ``` -- Prune filters using [FPGM](https://arxiv.org/abs/1811.00250) algorithm: +- 使用 [FPGM](https://arxiv.org/abs/1811.00250) 算法剪裁: ```shell --prune_criterion=geometry_median ``` -## Upcoming +## 敬请关注后续发布 -- [ ] More models configurations -- [ ] Quantized models +- [ ] 更多模型 +- [ ] 量化模型 diff --git a/configs/mobile/README_en.md b/configs/mobile/README_en.md new file mode 100755 index 000000000..0108e8c67 --- /dev/null +++ b/configs/mobile/README_en.md @@ -0,0 +1,82 @@ +English | [简体中文](README.md) + +# Mobile Model Zoo + + +## Models + +This directory contains models optimized for mobile applications, at present the following models included: + +| Backbone | Architecture | Input | Image/gpu 1 | Lr schd | Box AP | Download 2 | +|--------------------------|---------------------------|-------|------------------------|---------------|--------|-----------------------| +| MobileNetV3 Small | SSDLite | 320 | 64 | 400K (cosine) | 16.6 | [Link](https://paddlemodels.bj.bcebos.com/object_detection/ssdlite_mobilenet_v3_small.tar.gz) | +| MobileNetV3 Large | SSDLite | 320 | 64 | 400K (cosine) | 22.8 | [Link](https://paddlemodels.bj.bcebos.com/object_detection/ssdlite_mobilenet_v3_large.tar.gz) | +| MobileNetV3 Large w/ FPN | Cascade RCNN | 320 | 2 | 500k (cosine) | 25.0 | [Link](https://paddlemodels.bj.bcebos.com/object_detection/cascade_rcnn_mobilenetv3_fpn_320.tar.gz) | +| MobileNetV3 Large w/ FPN | Cascade RCNN | 640 | 2 | 500k (cosine) | 30.2 | [Link](https://paddlemodels.bj.bcebos.com/object_detection/cascade_rcnn_mobilenetv3_fpn_640.tar.gz) | +| MobileNetV3 Large | YOLOv3 | 320 | 8 | 500K | 27.1 | [Link](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v3.tar.gz) | +| MobileNetV3 Large | YOLOv3 Prune 3 | 320 | 8 | - | 24.6 | [Link](https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v3_prune86_FPGM_320.tar.gz) | + +**Notes**: + +- [1] All models are trained on 8 GPUs. +- [2] Each tarball contains the following files + - model weight file (`.pdparams` or `.tar`) + - inference model files (`__model__` and `__params__`) + - Paddle-Lite model file (`.nb`) +- [3] See the note section on how YOLO head is pruned + + +## Benchmarks Results + +- Models are benched on following chipsets with [Paddle-Lite](https://github.com/PaddlePaddle/Paddle-Lite) 2.6 (to be released) + - Qualcomm Snapdragon 625 + - Qualcomm Snapdragon 835 + - Qualcomm Snapdragon 845 + - Qualcomm Snapdragon 855 + - HiSilicon Kirin 970 + - HiSilicon Kirin 980 +- With 1 CPU thread (latency numbers are in ms) + + | | SD625 | SD835 | SD845 | SD855 | Kirin 970 | Kirin 980 | + |------------------|---------|---------|---------|---------|-----------|-----------| + | SSDLite Large | 289.071 | 134.408 | 91.933 | 48.2206 | 144.914 | 55.1186 | + | SSDLite Small | 122.932 | 57.1914 | 41.003 | 22.0694 | 61.5468 | 25.2106 | + | YOLOv3 baseline | 1082.5 | 435.77 | 317.189 | 155.948 | 536.987 | 178.999 | + | YOLOv3 prune | 253.98 | 131.279 | 89.4124 | 48.2856 | 122.732 | 55.8626 | + | Cascade RCNN 320 | 286.526 | 125.635 | 87.404 | 46.184 | 149.179 | 52.9994 | + | Cascade RCNN 640 | 1115.66 | 495.926 | 351.361 | 189.722 | 573.558 | 207.917 | +- With 4 CPU threads (latency numbers are in ms) + + | | SD625 | SD835 | SD845 | SD855 | Kirin 970 | Kirin 980 | + |------------------|---------|---------|---------|---------|-----------|-----------| + | SSDLite Large | 107.535 | 51.1382 | 34.6392 | 20.4978 | 50.5598 | 24.5318 | + | SSDLite Small | 51.5704 | 24.5156 | 18.5486 | 11.4218 | 24.9946 | 16.7158 | + | YOLOv3 baseline | 413.486 | 184.248 | 133.624 | 75.7354 | 202.263 | 126.435 | + | YOLOv3 prune | 98.5472 | 53.6228 | 34.4306 | 21.3112 | 44.0722 | 31.201 | + | Cascade RCNN 320 | 131.515 | 59.6026 | 39.4338 | 23.5802 | 58.5046 | 36.9486 | + | Cascade RCNN 640 | 473.083 | 224.543 | 156.205 | 100.686 | 231.108 | 138.391 | + + +## Notes on YOLOv3 pruning + +We pruned the YOLO-head and distill the pruned model with YOLOv3-ResNet34 as the teacher, which has a higher mAP on COCO (31.4 with 320\*320 input). + +The following configurations can be used for pruning: + +- Prune with fixed ratio, overall prune ratios is 86% + + ```shell + --pruned_params="yolo_block.0.0.0.conv.weights,yolo_block.0.0.1.conv.weights,yolo_block.0.1.0.conv.weights,yolo_block.0.1.1.conv.weights,yolo_block.0.2.conv.weights,yolo_block.0.tip.conv.weights,yolo_block.1.0.0.conv.weights,yolo_block.1.0.1.conv.weights,yolo_block.1.1.0.conv.weights,yolo_block.1.1.1.conv.weights,yolo_block.1.2.conv.weights,yolo_block.1.tip.conv.weights,yolo_block.2.0.0.conv.weights,yolo_block.2.0.1.conv.weights,yolo_block.2.1.0.conv.weights,yolo_block.2.1.1.conv.weights,yolo_block.2.2.conv.weights,yolo_block.2.tip.conv.weights" \ + --pruned_ratios="0.75,0.75,0.75,0.75,0.75,0.75,0.75,0.75,0.75,0.75,0.75,0.75,0.875,0.875,0.875,0.875,0.875,0.875" + ``` +- Prune filters using [FPGM](https://arxiv.org/abs/1811.00250) algorithm: + + ```shell + --prune_criterion=geometry_median + ``` + + +## Upcoming + +- [ ] More models configurations +- [ ] Quantized models -- GitLab