下载model——更名为model_8.1:
sudo apt install git
使用码云复制github项目,然后从码云下载
git clone http...
按照官方指导步骤:
install点击打开链接
config点击打开链接:
配置文件从samples中复制faster_rcnn_resnet101_voc07.config
prepare点击打开链接
running点击打开链接:
新建的modles下面的modle文件命名为faster_rcnn_resnet101
上述步骤中:
配置文件faster_rcnn_resnet101_voc07.config和运行代码中的路径为:
modle zoo:
/home/zhanghao/models_8.1/research/faster_rcnn_resnet101_coco_2018_01_28
record文件:
/home/zhanghao/models_8.1/research/data
运行代码:
python3 object_detection/train.py \--logtostderr \
--pipeline_config_path=/home/zhanghao/models_8.1/research/models/faster_rcnn_resnet101/faster_rcnn_resnet101_voc07.config \
--train_dir=/home/zhanghao/models_8.1/research/models/faster_rcnn_resnet101/train/
程序中需要把learning_shedules.py文件的 第167-169行由
rate_index = tf.reduce_max(tf.where(tf.greater_equal(global_step, boundaries),
range(num_boundaries),
[0] * num_boundaries))
改为
rate_index = tf.reduce_max(tf.where(tf.greater_equal(global_step, boundaries),
list(range(num_boundaries)),
[0] * num_boundaries))