RuntimeError：启用急切执行时不支持 tf.metrics.mean_iou

Question

我正在尝试获取两组边界框之间的平均 IOU 指标。我遇到了这个错误：RuntimeError：启用急切执行时不支持 tf.metrics.mean_iou。抛出错误的可重现代码：import tensorflow as tfa = [[0.462, 0.465, 0.492, 0.58],     [0.435, 0.385, 0.462, 0.467],     [0.586, 0.449, 0.627, 0.616],     [0.431, 0.671, 0.494, 0.8],     [0.519, 0.282, 0.586, 0.449]]b = [[0.465, 0.467, 0.491, 0.586],     [0.432, 0.488, 0.464, 0.59],     [0.585, 0.442, 0.633, 0.625],     [0.429, 0.664, 0.493, 0.805],     [0.524, 0.502, 0.594, 0.642]]tf.compat.v1.metrics.mean_iou(a, b, num_classes=1)我有什么选择？

Answer 1

我能够让它发挥作用：

import tensorflow as tf

a = [[0.462, 0.465, 0.492, 0.58],

     [0.435, 0.385, 0.462, 0.467],

     [0.586, 0.449, 0.627, 0.616],

     [0.431, 0.671, 0.494, 0.8],

     [0.519, 0.282, 0.586, 0.449]]

b = [[0.465, 0.467, 0.491, 0.586],

     [0.432, 0.488, 0.464, 0.59],

     [0.585, 0.442, 0.633, 0.625],

     [0.429, 0.664, 0.493, 0.805],

     [0.524, 0.502, 0.594, 0.642]]

metric = tf.keras.metrics.MeanIoU(num_classes=1)

print(metric(a, b))

>> tf.Tensor(1.0, shape=(), dtype=float32)