引言

基于PP-YOLOv2训练完模型后,The training process can refer to the author's previous blog【PP-YOLOv2】训练自定义的数据集,The next step is to test.

1. 评估

CUDA_VISIBLE_DEVICES=5 python tools/eval.py -c configs/ppyolo/ppyolov2_r50vd_dcn_voc.yml -o weights=/data3/205B/PaddleDetection/output/ppyolov2_r50vd_dcn_voc/model_final --output_eval=/data3/205B/PaddleDetection/output/eval 

并对eval.py进行如下修改：

After executing the command, the terminal displays the following：

各类的PRCurves are saved in the project root directoryvoc_pr_curve folder文件夹下：

2. 测试数据（推理）

参考官方文档：PP-YOLO 模型
执行代码如下：

 1. tools/infer.py  # 运行测试代码
 2. configs/ppyolo/ppyolov2_r50vd_dcn_voc.yml   # 配置文件
 3. weights=/data3/205B/PaddleDetection/output/ppyolov2_r50vd_dcn_voc/model_final  # 训练得到的权重
 4. infer_dir=/data3/205B/PaddleDetection/dataset/Track_fasteners/images  # 测试图片路径
 5. output_dir=/data3/205B/PaddleDetection/output/test_out  # 输出路径

# Saved with trainingcheckpoint
CUDA_VISIBLE_DEVICES=5 python tools/infer.py -c configs/ppyolo/ppyolov2_r50vd_dcn_voc.yml -o weights=/data3/205B/PaddleDetection/output/ppyolov2_r50vd_dcn_voc/model_final --infer_dir=/data3/205B/PaddleDetection/dataset/Track_fasteners/images --output_dir=/data3/205B/PaddleDetection/output/test_out

Finally, the result image is obtained at the specified path,Here the author's output path is：/data3/205B/PaddleDetection/output/test_out
The resulting image is shown below：