逆风者

１.搭建faster_rcnn_inception_resnet_v2_atrous_lowproposals_coco时遇到的问题（1）没有同名配置文件问题解决办法：从models/research/object_detection/samples/config下复制faster_rcnn_inception_resnet_v2_atrous_coco.config，参考已下载的预训...

1、num_examples：想要验证的图片数量      max_evals：验证循环次数/home/ubuntu/GoogleAPI/models/research/object_detection      metrics_set：选择验证的方式，有以下几个选项（在research/object_detection/legacy/evaluator.py中）：eval_co...

跑隐含模型，其预训练模型并没有在官网公布，需要重头开始训练步骤：1、复制同类配置文件，修改其中的特征提取网络名称，并注释掉加载预训练权重的代码2、在python安装目录下的object_detection下，修改builders/model_builder.py（我的目录在/home/ubuntu/anaconda3/lib/python3.6/site-packages/...

1、trainer.py（1）tf.train.Saver()函数：保存的检查点文件个数tf.train.Saver.__init__(var_list=None, reshape=False, sharded=False, max_to_keep=5, keep_checkpoint_every_n_hours=10000.0, name=None, restore_se...

一、TensorFlow Object Detection API的安装见本人以下两篇博客：windows10下安装TensorFlow Object Detection API：https://mp.youkuaiyun.com/postedit/82023949Ubuntu18.04下安装TensorFlow Object Detection API：https://mp.youkuaiyun.com/pos...

相关代码在tensorflow源码中：site-packages\tensorflow\contrib\slim\python\slim\learning.py中def train()--------->负责程序运行整体流程... logging.info('Starting Session.') ... logging.info...

'DetectionBoxes_Precision/mAP': mean average precision over classes averaged over IOU thresholds ranging from .5 to .95 with .05 increments.'DetectionBoxes_Precision/mAP@.50IOU': mean average precis...