运行eval_voc.py之后，output文件为空

[wasedd]

下面是运行eval_voc.py之后的结果，

生成output文件，但内容为空。没有测试结果的图片。请问是什么问题，该怎么解决？ ps：yolov4模型，无注意力模块 参数设置如下图：

还有，得到的预测结果pred_result如下，依然存在较多候选框，似乎nms没有起到筛选作用。这如何解决？

[Mahmood-Hussain]

Hello,

1. Create a directory detection_result inside your project directory
2. add --mode eval to your command for getting results. So, the command will look like this CUDA_VISIBLE_DEVICES=0 python3 eval_voc.py --weight_path weights/best.pt --gpu_id 0 --visiual data/ValImages --mode eval change --visiual to your image directory for test

[wasedd]

Hello,

1. Create a directory detection_result inside your project directory
2. add --mode eval to your command for getting results. So, the command will look like this CUDA_VISIBLE_DEVICES=0 python3 eval_voc.py --weight_path weights/best.pt --gpu_id 0 --visiual data/ValImages --mode eval change --visiual to your image directory for test

非常感谢。

thank you very much！

[wasedd]

Hello,

1. Create a directory detection_result inside your project directory
2. add --mode eval to your command for getting results. So, the command will look like this CUDA_VISIBLE_DEVICES=0 python3 eval_voc.py --weight_path weights/best.pt --gpu_id 0 --visiual data/ValImages --mode eval change --visiual to your image directory for test

大佬，运行eval_voc.py之后，得到的 预测结果 pred_result文件中，为什么 同一张图像对应多个预测框，虽然是按照得分从最高到最低排列。怎么样可以得到一对一的结果。谢谢。

[Mahmood-Hussain]

Hello, I do not understand Chinese after a quick google translation I am replying to what I understood.

Some of the results will be discarded which are below IoU threshold. In other words bounding boxes with low probability will be discarded so you do not need to worry the result is fine

[wasedd]

Hello, I do not understand Chinese after a quick google translation I am replying to what I understood.

Some of the results will be discarded which are below IoU threshold. In other words bounding boxes with low probability will be discarded so you do not need to worry the result is fine

Hello, brother.Thank you very much for answering my question. As you said，I konwn，some of the results will be discarded which are below IoU threshold. Because what I get after running eval_voc.py is this result. And according to this result, it is obvious that it cannot run successfully in get_map.py.So I have another questions.

1. Isn't the result obtained by running this program the final prediction result ？ 2 .Or do I need to take another step after I get it to screen out the final appropriate prediction results？ Thanks again, brother.

[lifafang]

https://github.com/argusswift/YOLOv4-pytorch/issues/201#issuecomment-885987541 可以留一下您的联系方式，讨论一下这个问题吗？ @wasedd

[wasedd]

#201 (comment) 可以留一下您的联系方式，讨论一下这个问题吗？ @wasedd

谢谢回答。wx：xiazheyu12