我正在尝试使用Tensorflow对象检测API和我的自定义数据集进行训练和对象检测。当我使用我的tfrecord文件进行训练时，我看起来像这样：...
INFO:tensorflow:Recording summary at step 0.
INFO:tensorflow:global step 1: loss = 0.0340 (4.919 sec/step)
INFO:tensorflow:global step 2: loss = 0.0340 (0.202 sec/step)
INFO:tensorflow:global step 3: loss = 72.7242 (0.216 sec/step)
INFO:tensorflow:global step 4: loss = 0.0340 (0.206 sec/step)
INFO:tensorflow:global step 5: loss = 0.0342 (0.191 sec/step)
INFO:tensorflow:global step 6: loss = 0.0344 (0.204 sec/step)
INFO:tensorflow:global step 7: loss = 0.0347 (0.200 sec/step)
INFO:tensorflow:global step 8: loss = 0.0349 (0.207 sec/step)
INFO:tensorflow:global step 9: loss = 0.0352 (0.184 sec/step)
INFO:tensorflow:global step 10: loss = 0.0355 (0.216 sec/step)
INFO:tensorflow:global step 11: loss = 0.0358 (0.199 sec/step)
INFO:tensorflow:global step 12: loss = 44.0913 (0.237 sec/step)
INFO:tensorflow:global step 13: loss = 0.0365 (0.230 sec/step)
INFO:tensorflow:global step 14: loss = 0.0368 (0.214 sec/step)
INFO:tensorflow:global step 15: loss = 0.0372 (0.227 sec/step)
当我冻结模型并尝试测试推理时，它完全无法生成任何有意义的边界框。
到目前为止，我尝试了很多东西。我手动检查了我的tfrecord文件的内容（意味着我从我的tfrecord中提取了二进制jpegs和边界框以显示它们）并且它们看起来很好。
边界框正确放置在目标顶部，训练图像和bbox看起来很好。我不确定是什么导致这种错误，特别是考虑到我已经检查了我的自定义tfrecord文件。
感谢任何帮助。