tensorflow.config
1、eval_config下的num_examples 的含义(类似ssd_mobilenet_v1_coco.config中的配置)
个人比较倾向于下面这种解释,但是 没有找到官方的说法:(What does _num_examples_ 2000_ mean in TensorFlow object detection config file_ - Stack Overflow.html (https://stackoverflow.com/questions/47086630/what-does-num-examples-2000-mean-in-tensorflow-object-detection-config-file))
Actually this means only the same top num_examples samples in your evaluation dataset will be used in each run of evaluation.
2、资料:
TensorFlow Object Detection API配置文件解读_人工智能_逆风者-CSDN博客.html(https://blog.csdn.net/qq_28019591/article/details/86560116)
tensorflow训练模型配置文件详解 神狗.html(http://godgou.com/blog/articles/433.html)
ZC:文章里什么意思?是通过 proto文件里面的 成员变量名及其默认值 来配置config文件中的对应项?
使用 Tensorflow Object Detection API 训练自己的模型.html(https://laddiexu.github.io/tech/2017/11/04/TF-ODYourData.html)
ZC:评估模型:python object_detection/eval.py --logtostderr --pipeline_config_path=object_detection/training/ssd_mobilenet_v1_toothpaste.config --checkpoint_dir=object_detection/traininglog --eval_dir=object_detection/eveldire
谷歌开放的TensorFlow Object Detection API 效果如何?对业界有什么影响? - 知乎.html(https://www.zhihu.com/question/61173908)
ZC:一些资料地址 和 一些概述。2017"知乎"上的,还是可以一看的
3、对 config文件中 eval_config -->num_examples的解释是这样的:
# (Optional): Set this to the number of images in your <PATH_TO_IMAGES_FOLDER>/train # if you want to also run evaluation
来自于:(看网址,感觉下面2个东西是一个出处)
Training Custom Object Detector — TensorFlow Object Detection API tutorial documentation.html(https://tensorflow-object-detection-api-tutorial.readthedocs.io/en/latest/training.html)
tensorflow-object-detection-api-tutorial-readthedocs-io-en-latest.pdf(https://readthedocs.org/projects/tensorflow-object-detection-api-tutorial/downloads/pdf/latest/)
上面2个文章中看起来 对 eval_config -->num_examples的讲解还蛮正规的,但是 用“<PATH_TO_IMAGES_FOLDER>/train”的数量 总觉得不对劲... 不知道官方文档里面是怎么样解释的...
4、
5、