更换文档检测模型

paddle_detection/configs/ppvehicle/vehicle_yolov3/README.md

@@ -0,0 +1,53 @@
+English | [简体中文](README_cn.md)
+# PaddleDetection applied for specific scenarios
+
+We provide some models implemented by PaddlePaddle to detect objects in specific scenarios, users can download the models and use them in these scenarios.
+
+| Task                 | Algorithm | Box AP | Download                                                                                | Configs |
+|:---------------------|:---------:|:------:| :-------------------------------------------------------------------------------------: |:------:|
+| Vehicle Detection    |  YOLOv3  |  54.5  | [model](https://paddledet.bj.bcebos.com/models/vehicle_yolov3_darknet.pdparams) | [config](./vehicle_yolov3_darknet.yml) |
+
+## Vehicle Detection
+
+One of major applications of vehichle detection is traffic monitoring. In this scenary, vehicles to be detected are mostly captured by the cameras mounted on top of traffic light columns.
+
+### 1. Network
+
+The network for detecting vehicles is YOLOv3, the backbone of which is Dacknet53.
+
+### 2. Configuration for training
+
+PaddleDetection provides users with a configuration file [yolov3_darknet53_270e_coco.yml](https://github.com/PaddlePaddle/PaddleDetection/blob/develop/configs/yolov3/yolov3_darknet53_270e_coco.yml) to train YOLOv3 on the COCO dataset, compared with this file, we modify some parameters as followed to conduct the training for vehicle detection:
+
+* num_classes: 6
+* anchors: [[8, 9], [10, 23], [19, 15], [23, 33], [40, 25], [54, 50], [101, 80], [139, 145], [253, 224]]
+* nms/nms_top_k: 400
+* nms/score_threshold: 0.005
+* dataset_dir: dataset/vehicle
+
+### 3. Accuracy
+
+The accuracy of the model trained and evaluated on our private data is shown as followed:
+
+AP at IoU=.50:.05:.95 is 0.545.
+
+AP at IoU=.50 is 0.764.
+
+### 4. Inference
+
+Users can employ the model to conduct the inference:
+
+```
+export CUDA_VISIBLE_DEVICES=0
+python -u tools/infer.py -c configs/ppvehicle/vehicle_yolov3/vehicle_yolov3_darknet.yml \
+                         -o weights=https://paddledet.bj.bcebos.com/models/vehicle_yolov3_darknet.pdparams \
+                         --infer_dir configs/ppvehicle/vehicle_yolov3/demo \
+                         --draw_threshold 0.2 \
+                         --output_dir configs/ppvehicle/vehicle_yolov3/demo/output
+```
+
+Some inference results are visualized below:
+
+![](../../../docs/images/VehicleDetection_001.jpeg)
+
+![](../../../docs/images/VehicleDetection_005.png)

paddle_detection/configs/ppvehicle/vehicle_yolov3/README_cn.md

@@ -0,0 +1,54 @@
+[English](README.md) | 简体中文
+# 特色垂类检测模型
+
+我们提供了针对不同场景的基于PaddlePaddle的检测模型，用户可以下载模型进行使用。
+
+| 任务                 | 算法 | 精度(Box AP) | 下载                                                                                | 配置文件 |
+|:---------------------|:---------:|:------:| :---------------------------------------------------------------------------------: | :------:|
+| 车辆检测    |  YOLOv3  |  54.5  | [下载链接](https://paddledet.bj.bcebos.com/models/vehicle_yolov3_darknet.pdparams) | [配置文件](./vehicle_yolov3_darknet.yml) |
+
+
+## 车辆检测（Vehicle Detection）
+
+车辆检测的主要应用之一是交通监控。在这样的监控场景中，待检测的车辆多为道路红绿灯柱上的摄像头拍摄所得。
+
+### 1. 模型结构
+
+Backbone为Dacknet53的YOLOv3。
+
+### 2. 训练参数配置
+
+PaddleDetection提供了使用COCO数据集对YOLOv3进行训练的参数配置文件[yolov3_darknet53_270e_coco.yml](https://github.com/PaddlePaddle/PaddleDetection/blob/develop/configs/yolov3/yolov3_darknet53_270e_coco.yml)，与之相比，在进行车辆检测的模型训练时，我们对以下参数进行了修改：
+
+* num_classes: 6
+* anchors: [[8, 9], [10, 23], [19, 15], [23, 33], [40, 25], [54, 50], [101, 80], [139, 145], [253, 224]]
+* nms/nms_top_k: 400
+* nms/score_threshold: 0.005
+* dataset_dir: dataset/vehicle
+
+### 3. 精度指标
+
+模型在我们内部数据上的精度指标为：
+
+IOU=.50:.05:.95时的AP为 0.545。
+
+IOU=.5时的AP为 0.764。
+
+### 4. 预测
+
+用户可以使用我们训练好的模型进行车辆检测：
+
+```
+export CUDA_VISIBLE_DEVICES=0
+python -u tools/infer.py -c configs/ppvehicle/vehicle_yolov3/vehicle_yolov3_darknet.yml \
+                         -o weights=https://paddledet.bj.bcebos.com/models/vehicle_yolov3_darknet.pdparams \
+                         --infer_dir configs/ppvehicle/vehicle_yolov3/demo \
+                         --draw_threshold 0.2 \
+                         --output_dir configs/ppvehicle/vehicle_yolov3/demo/output
+```
+
+预测结果示例：
+
+![](../../../docs/images/VehicleDetection_001.jpeg)
+
+![](../../../docs/images/VehicleDetection_005.png)

paddle_detection/configs/ppvehicle/vehicle_yolov3/vehicle_yolov3_darknet.yml

@@ -0,0 +1,42 @@
+_BASE_: [
+  '../../datasets/coco_detection.yml',
+  '../../runtime.yml',
+  '../../yolov3/_base_/optimizer_270e.yml',
+  '../../yolov3/_base_/yolov3_darknet53.yml',
+  '../../yolov3/_base_/yolov3_reader.yml',
+]
+
+snapshot_epoch: 5
+weights: https://paddledet.bj.bcebos.com/models/vehicle_yolov3_darknet.pdparams
+
+YOLOv3Head:
+  anchors: [[8, 9], [10, 23], [19, 15],
+            [23, 33], [40, 25], [54, 50],
+            [101, 80], [139, 145], [253, 224]]
+
+BBoxPostProcess:
+  nms:
+    name: MultiClassNMS
+    keep_top_k: 100
+    score_threshold: 0.005
+    nms_threshold: 0.45
+    nms_top_k: 400
+
+num_classes: 6
+
+TrainDataset:
+  !COCODataSet
+    dataset_dir: dataset/vehicle
+    anno_path: annotations/instances_train2017.json
+    image_dir: train2017
+    data_fields: ['image', 'gt_bbox', 'gt_class', 'is_crowd']
+
+EvalDataset:
+  !COCODataSet
+    dataset_dir: dataset/vehicle
+    anno_path: annotations/instances_val2017.json
+    image_dir: val2017
+
+TestDataset:
+  !ImageFolder
+    anno_path: configs/ppvehicle/vehicle_yolov3/vehicle.json