Tensorflow object detection API之faster_rcnn_resnet101_coco_2018_1_28 with TF 1.7出錯

當訓練faster_rcnn_resnet101_coco_2018_1_28時，運行如下命令

python object_detection/train.py

--logtostderr

--pipeline_config_path=${PATH_TO_YOUR_PIPELINE_CONFIG}

--train_dir=${PATH_TO_TRAIN_DIR}

出現錯誤：

ValueError: Tried to convert t to a tensor and failed. Error: Argument must be a dense tensor: range(0, 3) - got shape [3], but wanted []

出錯原因：python3的兼容問題

解決辦法：把research/object_detection/utils/learning_schedules.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))

改為

rate_index = tf.reduce_max(tf.where(tf.greater_equal(global_step, boundaries),

list(range(num_boundaries)),

[0] * num_boundaries))

推薦閱讀：