更换文档检测模型

2024-08-27 14:42:45 +08:00
parent aea6f19951
commit 1514e09c40
2072 changed files with 254336 additions and 4967 deletions
--- a/paddle_detection/configs/rotate/s2anet/README.md
+++ b/paddle_detection/configs/rotate/s2anet/README.md
@@ -0,0 +1,104 @@
+简体中文 | [English](README_en.md)
+
+# S2ANet
+
+## 内容
+- [简介](#简介)
+- [模型库](#模型库)
+- [使用说明](#使用说明)
+- [预测部署](#预测部署)
+- [引用](#引用)
+
+## 简介
+
+[S2ANet](https://arxiv.org/pdf/2008.09397.pdf)是用于检测旋转框的模型.
+
+## 模型库
+
+| 模型 | Conv类型 | mAP | 学习率策略 | 角度表示 | 数据增广 | GPU数目 | 每GPU图片数目 | 模型下载 | 配置文件 |
+|:---:|:------:|:----:|:---------:|:-----:|:--------:|:-----:|:------------:|:-------:|:------:|
+| S2ANet | Conv | 71.45 | 2x | le135 | - | 4 | 2 | [model](https://paddledet.bj.bcebos.com/models/s2anet_conv_2x_dota.pdparams) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/rotate/s2anet/s2anet_conv_2x_dota.yml) |
+| S2ANet | AlignConv | 73.84 | 2x | le135 | - | 4 | 2 | [model](https://paddledet.bj.bcebos.com/models/s2anet_alignconv_2x_dota.pdparams) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/rotate/s2anet/s2anet_alignconv_2x_dota.yml) |
+
+**注意：**
+
+- 如果**GPU卡数**或者**batch size**发生了改变，你需要按照公式 **lr<sub>new</sub> = lr<sub>default</sub> * (batch_size<sub>new</sub> * GPU_number<sub>new</sub>) / (batch_size<sub>default</sub> * GPU_number<sub>default</sub>)** 调整学习率。
+- 模型库中的模型默认使用单尺度训练单尺度测试。如果数据增广一栏标明MS，意味着使用多尺度训练和多尺度测试。如果数据增广一栏标明RR，意味着使用RandomRotate数据增广进行训练。
+- 这里使用`multiclass_nms`，与原作者使用nms略有不同。
+
+
+## 使用说明
+
+参考[数据准备](../README.md#数据准备)准备数据。
+
+### 1. 训练
+
+GPU单卡训练
+```bash
+export CUDA_VISIBLE_DEVICES=0
+python tools/train.py -c configs/rotate/s2anet/s2anet_1x_spine.yml
+```
+
+GPU多卡训练
+```bash
+export CUDA_VISIBLE_DEVICES=0,1,2,3
+python -m paddle.distributed.launch --gpus 0,1,2,3 tools/train.py -c configs/rotate/s2anet/s2anet_1x_spine.yml
+```
+
+可以通过`--eval`开启边训练边测试。
+
+### 2. 评估
+```bash
+python tools/eval.py -c configs/rotate/s2anet/s2anet_1x_spine.yml -o weights=output/s2anet_1x_spine/model_final.pdparams
+
+# 使用提供训练好的模型评估
+python tools/eval.py -c configs/rotate/s2anet/s2anet_1x_spine.yml -o weights=https://paddledet.bj.bcebos.com/models/s2anet_1x_spine.pdparams
+```
+
+### 3. 预测
+执行如下命令，会将图像预测结果保存到`output`文件夹下。
+```bash
+python tools/infer.py -c configs/rotate/s2anet/s2anet_1x_spine.yml -o weights=output/s2anet_1x_spine/model_final.pdparams --infer_img=demo/39006.jpg --draw_threshold=0.3
+```
+使用提供训练好的模型预测：
+```bash
+python tools/infer.py -c configs/rotate/s2anet/s2anet_1x_spine.yml -o weights=https://paddledet.bj.bcebos.com/models/s2anet_1x_spine.pdparams --infer_img=demo/39006.jpg --draw_threshold=0.3
+```
+
+### 4. DOTA数据评估
+执行如下命令，会在`output`文件夹下将每个图像预测结果保存到同文件夹名的txt文本中。
+```
+python tools/infer.py -c configs/rotate/s2anet/s2anet_alignconv_2x_dota.yml -o weights=https://paddledet.bj.bcebos.com/models/s2anet_alignconv_2x_dota.pdparams --infer_dir=/path/to/test/images --output_dir=output --visualize=False --save_results=True
+```
+参考[DOTA Task](https://captain-whu.github.io/DOTA/tasks.html), 评估DOTA数据集需要生成一个包含所有检测结果的zip文件，每一类的检测结果储存在一个txt文件中，txt文件中每行格式为：`image_name score x1 y1 x2 y2 x3 y3 x4 y4`。将生成的zip文件提交到[DOTA Evaluation](https://captain-whu.github.io/DOTA/evaluation.html)的Task1进行评估。你可以执行以下命令生成评估文件
+```
+python configs/rotate/tools/generate_result.py --pred_txt_dir=output/ --output_dir=submit/ --data_type=dota10
+
+zip -r submit.zip submit
+```
+
+## 预测部署
+
+Paddle中`multiclass_nms`算子的输入支持四边形输入，因此部署时可以不需要依赖旋转框IOU计算算子。
+
+部署教程请参考[预测部署](../../../deploy/README.md)
+
+
+## 引用
+```
+@article{han2021align,  
+  author={J. {Han} and J. {Ding} and J. {Li} and G. -S. {Xia}},  
+  journal={IEEE Transactions on Geoscience and Remote Sensing},  
+  title={Align Deep Features for Oriented Object Detection},  
+  year={2021},
+  pages={1-11},  
+  doi={10.1109/TGRS.2021.3062048}}
+
+@inproceedings{xia2018dota,
+  title={DOTA: A large-scale dataset for object detection in aerial images},
+  author={Xia, Gui-Song and Bai, Xiang and Ding, Jian and Zhu, Zhen and Belongie, Serge and Luo, Jiebo and Datcu, Mihai and Pelillo, Marcello and Zhang, Liangpei},
+  booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
+  pages={3974--3983},
+  year={2018}
+}
+```
--- a/paddle_detection/configs/rotate/s2anet/README_en.md
+++ b/paddle_detection/configs/rotate/s2anet/README_en.md
@@ -0,0 +1,102 @@
+English | [简体中文](README.md)
+
+# S2ANet
+
+## Content
+- [Introduction](#Introduction)
+- [Model Zoo](#Model-Zoo)
+- [Getting Start](#Getting-Start)
+- [Deployment](#Deployment)
+- [Citations](#Citations)
+
+## Introduction
+
+[S2ANet](https://arxiv.org/pdf/2008.09397.pdf) is used to detect rotated objects.
+
+## Model Zoo
+| Model | Conv Type | mAP | Lr Scheduler | Angle | Aug | GPU Number | images/GPU | download | config |
+|:---:|:------:|:----:|:---------:|:-----:|:--------:|:-----:|:------------:|:-------:|:------:|
+| S2ANet | Conv | 71.45 | 2x | le135 | - | 4 | 2 | [model](https://paddledet.bj.bcebos.com/models/s2anet_conv_2x_dota.pdparams) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/rotate/s2anet/s2anet_conv_2x_dota.yml) |
+| S2ANet | AlignConv | 73.84 | 2x | le135 | - | 4 | 2 | [model](https://paddledet.bj.bcebos.com/models/s2anet_alignconv_2x_dota.pdparams) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/rotate/s2anet/s2anet_alignconv_2x_dota.yml) |
+
+**Notes:**
+- if **GPU number** or **mini-batch size** is changed, **learning rate** should be adjusted according to the formula **lr<sub>new</sub> = lr<sub>default</sub> * (batch_size<sub>new</sub> * GPU_number<sub>new</sub>) / (batch_size<sub>default</sub> * GPU_number<sub>default</sub>)**.
+- Models in model zoo is trained and tested with single scale by default. If `MS` is indicated in the data augmentation column, it means that multi-scale training and multi-scale testing are used. If `RR` is indicated in the data augmentation column, it means that RandomRotate data augmentation is used for training.
+- `multiclass_nms` is used here, which is slightly different from the original author's use of NMS.
+
+## Getting Start
+
+Refer to [Data-Preparation](../README_en.md#Data-Preparation) to prepare data.
+
+### 1. Train
+
+Single GPU Training
+```bash
+export CUDA_VISIBLE_DEVICES=0
+python tools/train.py -c configs/rotate/s2anet/s2anet_1x_spine.yml
+```
+
+Multiple GPUs Training
+```bash
+export CUDA_VISIBLE_DEVICES=0,1,2,3
+python -m paddle.distributed.launch --gpus 0,1,2,3 tools/train.py -c configs/rotate/s2anet/s2anet_1x_spine.yml
+```
+
+You can use `--eval`to enable train-by-test.
+
+### 2. Evaluation
+```bash
+python tools/eval.py -c configs/rotate/s2anet/s2anet_1x_spine.yml -o weights=output/s2anet_1x_spine/model_final.pdparams
+
+# Use a trained model to evaluate
+python tools/eval.py -c configs/rotate/s2anet/s2anet_1x_spine.yml -o weights=https://paddledet.bj.bcebos.com/models/s2anet_1x_spine.pdparams
+```
+
+### 3. Prediction
+Executing the following command will save the image prediction results to the `output` folder.
+```bash
+python tools/infer.py -c configs/rotate/s2anet/s2anet_1x_spine.yml -o weights=output/s2anet_1x_spine/model_final.pdparams --infer_img=demo/39006.jpg --draw_threshold=0.3
+```
+Prediction using models that provide training:
+```bash
+python tools/infer.py -c configs/rotate/s2anet/s2anet_1x_spine.yml -o weights=https://paddledet.bj.bcebos.com/models/s2anet_1x_spine.pdparams --infer_img=demo/39006.jpg --draw_threshold=0.3
+```
+
+### 4. DOTA Data evaluation
+Execute the following command, will save each image prediction result in `output` folder txt text with the same folder name.
+```
+python tools/infer.py -c configs/rotate/s2anet/s2anet_alignconv_2x_dota.yml -o weights=https://paddledet.bj.bcebos.com/models/s2anet_alignconv_2x_dota.pdparams --infer_dir=/path/to/test/images --output_dir=output --visualize=False --save_results=True
+```
+Refering to [DOTA Task](https://captain-whu.github.io/DOTA/tasks.html), You need to submit a zip file containing results for all test images for evaluation. The detection results of each category are stored in a txt file, each line of which is in the following format
+`image_id score x1 y1 x2 y2 x3 y3 x4 y4`. To evaluate, you should submit the generated zip file to the Task1 of [DOTA Evaluation](https://captain-whu.github.io/DOTA/evaluation.html). You can execute the following command to generate the file
+```
+python configs/rotate/tools/generate_result.py --pred_txt_dir=output/ --output_dir=submit/ --data_type=dota10
+
+zip -r submit.zip submit
+```
+
+## Deployment
+
+The inputs of the `multiclass_nms` operator in Paddle support quadrilateral inputs, so deployment can be done without relying on the rotating frame IOU operator.
+
+Please refer to the deployment tutorial[Predict deployment](../../../deploy/README_en.md)
+
+
+## Citations
+```
+@article{han2021align,  
+  author={J. {Han} and J. {Ding} and J. {Li} and G. -S. {Xia}},  
+  journal={IEEE Transactions on Geoscience and Remote Sensing},  
+  title={Align Deep Features for Oriented Object Detection},  
+  year={2021},
+  pages={1-11},  
+  doi={10.1109/TGRS.2021.3062048}}
+
+@inproceedings{xia2018dota,
+  title={DOTA: A large-scale dataset for object detection in aerial images},
+  author={Xia, Gui-Song and Bai, Xiang and Ding, Jian and Zhu, Zhen and Belongie, Serge and Luo, Jiebo and Datcu, Mihai and Pelillo, Marcello and Zhang, Liangpei},
+  booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
+  pages={3974--3983},
+  year={2018}
+}
+```
--- a/paddle_detection/configs/rotate/s2anet/_base_/s2anet.yml
+++ b/paddle_detection/configs/rotate/s2anet/_base_/s2anet.yml
@@ -0,0 +1,52 @@
+architecture: S2ANet
+pretrain_weights: https://paddledet.bj.bcebos.com/models/pretrained/ResNet50_vd_ssld_v2_pretrained.pdparams
+weights: output/s2anet_r50_fpn_1x_dota/model_final.pdparams
+
+
+# Model Achitecture
+S2ANet:
+  backbone: ResNet
+  neck: FPN
+  head: S2ANetHead
+
+ResNet:
+  depth: 50
+  variant: d
+  norm_type: bn
+  return_idx: [1,2,3]
+  num_stages: 4
+
+FPN:
+  in_channels: [256, 512, 1024]
+  out_channel: 256
+  spatial_scales: [0.25, 0.125, 0.0625]
+  has_extra_convs: True
+  extra_stage: 2
+  relu_before_extra_convs: False
+
+S2ANetHead:
+  anchor_strides: [8, 16, 32, 64, 128]
+  anchor_scales: [4]
+  anchor_ratios: [1.0]
+  anchor_assign: RBoxAssigner
+  stacked_convs: 2
+  feat_in: 256
+  feat_out: 256
+  align_conv_type: 'AlignConv'  # AlignConv Conv
+  align_conv_size: 3
+  use_sigmoid_cls: True
+  reg_loss_weight: [1.0, 1.0, 1.0, 1.0, 1.1]
+  cls_loss_weight: [1.1, 1.05]
+  nms_pre: 2000
+  nms:
+    name: MultiClassNMS
+    keep_top_k: -1
+    score_threshold: 0.05
+    nms_threshold: 0.1
+    normalized: False
+
+RBoxAssigner:
+  pos_iou_thr: 0.5
+  neg_iou_thr: 0.4
+  min_iou_thr: 0.0
+  ignore_iof_thr: -2
--- a/paddle_detection/configs/rotate/s2anet/_base_/s2anet_optimizer_1x.yml
+++ b/paddle_detection/configs/rotate/s2anet/_base_/s2anet_optimizer_1x.yml
@@ -0,0 +1,20 @@
+epoch: 12
+
+LearningRate:
+  base_lr: 0.01
+  schedulers:
+  - !PiecewiseDecay
+    gamma: 0.1
+    milestones: [7, 10]
+  - !LinearWarmup
+    start_factor: 0.3333333333333333
+    steps: 500
+
+OptimizerBuilder:
+  optimizer:
+    momentum: 0.9
+    type: Momentum
+  regularizer:
+    factor: 0.0001
+    type: L2
+  clip_grad_by_norm: 35
--- a/paddle_detection/configs/rotate/s2anet/_base_/s2anet_optimizer_2x.yml
+++ b/paddle_detection/configs/rotate/s2anet/_base_/s2anet_optimizer_2x.yml
@@ -0,0 +1,20 @@
+epoch: 24
+
+LearningRate:
+  base_lr: 0.01
+  schedulers:
+  - !PiecewiseDecay
+    gamma: 0.1
+    milestones: [14, 20]
+  - !LinearWarmup
+    start_factor: 0.3333333333333333
+    steps: 1000
+
+OptimizerBuilder:
+  optimizer:
+    momentum: 0.9
+    type: Momentum
+  regularizer:
+    factor: 0.0001
+    type: L2
+  clip_grad_by_norm: 35
--- a/paddle_detection/configs/rotate/s2anet/_base_/s2anet_reader.yml
+++ b/paddle_detection/configs/rotate/s2anet/_base_/s2anet_reader.yml
@@ -0,0 +1,44 @@
+worker_num: 4
+TrainReader:
+  sample_transforms:
+    - Decode: {}
+    - Poly2Array: {}
+    - RandomRFlip: {}
+    - RResize: {target_size: [1024, 1024], keep_ratio: True, interp: 2}
+    - Poly2RBox: {rbox_type: 'le135'}
+  batch_transforms:
+    - NormalizeImage: {mean: [0.485, 0.456, 0.406], std: [0.229, 0.224, 0.225], is_scale: True}
+    - Permute: {}
+    - PadRGT: {}
+    - PadBatch: {pad_to_stride: 32}
+  batch_size: 2
+  shuffle: true
+  drop_last: true
+
+
+EvalReader:
+  sample_transforms:
+    - Decode: {}
+    - Poly2Array: {}
+    - RResize: {target_size: [1024, 1024], keep_ratio: True, interp: 2}
+    - NormalizeImage: {mean: [0.485, 0.456, 0.406], std: [0.229, 0.224, 0.225], is_scale: True}
+    - Permute: {}
+  batch_transforms:
+    - PadBatch: {pad_to_stride: 32}
+  batch_size: 2
+  shuffle: false
+  drop_last: false
+  collate_batch: false
+
+
+TestReader:
+  sample_transforms:
+    - Decode: {}
+    - Resize: {interp: 2, target_size: [1024, 1024], keep_ratio: True}
+    - NormalizeImage: {is_scale: true, mean: [0.485,0.456,0.406], std: [0.229, 0.224,0.225]}
+    - Permute: {}
+  batch_transforms:
+    - PadBatch: {pad_to_stride: 32}
+  batch_size: 1
+  shuffle: false
+  drop_last: false
--- a/paddle_detection/configs/rotate/s2anet/s2anet_1x_spine.yml
+++ b/paddle_detection/configs/rotate/s2anet/s2anet_1x_spine.yml
@@ -0,0 +1,25 @@
+_BASE_: [
+  '../../datasets/spine_coco.yml',
+  '../../runtime.yml',
+  '_base_/s2anet_optimizer_1x.yml',
+  '_base_/s2anet.yml',
+  '_base_/s2anet_reader.yml',
+]
+
+weights: output/s2anet_1x_spine/model_final
+pretrain_weights: https://paddledet.bj.bcebos.com/models/s2anet_alignconv_2x_dota.pdparams
+
+# for 4 card
+LearningRate:
+  base_lr: 0.01
+  schedulers:
+  - !PiecewiseDecay
+    gamma: 0.1
+    milestones: [7, 10]
+  - !LinearWarmup
+    start_factor: 0.3333333333333333
+    epochs: 5
+
+S2ANetHead:
+  reg_loss_weight: [1.0, 1.0, 1.0, 1.0, 1.05]
+  cls_loss_weight: [1.05, 1.0]
--- a/paddle_detection/configs/rotate/s2anet/s2anet_alignconv_2x_dota.yml
+++ b/paddle_detection/configs/rotate/s2anet/s2anet_alignconv_2x_dota.yml
@@ -0,0 +1,10 @@
+_BASE_: [
+  '../../datasets/dota.yml',
+  '../../runtime.yml',
+  '_base_/s2anet_optimizer_2x.yml',
+  '_base_/s2anet.yml',
+  '_base_/s2anet_reader.yml',
+]
+pretrain_weights: https://paddledet.bj.bcebos.com/models/pretrained/ResNet50_vd_ssld_v2_pretrained.pdparams
+
+weights: output/s2anet_alignconv_2x_dota/model_final
--- a/paddle_detection/configs/rotate/s2anet/s2anet_conv_2x_dota.yml
+++ b/paddle_detection/configs/rotate/s2anet/s2anet_conv_2x_dota.yml
@@ -0,0 +1,19 @@
+_BASE_: [
+  '../../datasets/dota.yml',
+  '../../runtime.yml',
+  '_base_/s2anet_optimizer_2x.yml',
+  '_base_/s2anet.yml',
+  '_base_/s2anet_reader.yml',
+]
+weights: output/s2anet_conv_1x_dota/model_final
+pretrain_weights: https://paddledet.bj.bcebos.com/models/pretrained/ResNet50_cos_pretrained.pdparams
+
+ResNet:
+  depth: 50
+  variant: b
+  norm_type: bn
+  return_idx: [1,2,3]
+  num_stages: 4
+
+S2ANetHead:
+  align_conv_type: 'Conv'