Tensorflow object_detection 保存和加载微调模型的正确方法

Tensorflow object_detection correct way to save and load fine tune model

我正在使用 colabs 教程中的 this 示例来微调模型,训练后我想保存模型并加载到我的本地计算机上使用:

ckpt_manager = tf.train.CheckpointManager(ckpt, directory="test_data/checkpoint/", max_to_keep=5)
...
...
print('Done fine-tuning!')

ckpt_manager.save()
print('Checkpoint saved!')

但是在我的本地计算机上使用检查点文件恢复后没有检测到任何对象(分数太低)

我也试过

tf.saved_model.save(detection_model, '/content/new_model/')

并载入:

detection_model = tf.saved_model.load('/saved_model_20201226/')

input_tensor = tf.convert_to_tensor(image, dtype=tf.float32)
detections = detection_model(input_tensor)

给我这个错误: 类型错误:“_UserObject”对象不可调用

保存和加载微调模型的正确方法是什么?

编辑 1: 等待保存新的管道配置,之后终于成功了! 这是我的回答:

# Save new pipeline config
new_pipeline_proto = config_util.create_pipeline_proto_from_configs(configs)
config_util.save_pipeline_config(new_pipeline_proto, '/content/new_config')
exported_ckpt = tf.compat.v2.train.Checkpoint(model=detection_model)
ckpt_manager = tf.train.CheckpointManager(
exported_ckpt, directory="test_data/checkpoint/", max_to_keep=5)
...
...
print('Done fine-tuning!')

ckpt_manager.save()
print('Checkpoint saved!')

等待保存新的管道配置,之后终于成功了!这是我的回答:

# Save new pipeline config
new_pipeline_proto = config_util.create_pipeline_proto_from_configs(configs)
config_util.save_pipeline_config(new_pipeline_proto, '/content/new_config')

exported_ckpt = tf.compat.v2.train.Checkpoint(model=detection_model)
ckpt_manager = tf.train.CheckpointManager(
exported_ckpt, directory="test_data/checkpoint/", max_to_keep=5)
...
...
print('Done fine-tuning!')

ckpt_manager.save()
print('Checkpoint saved!')