CODEWITHSUNDEEP
tensorflowtensorflow-datasets
BadProgrammer
BadProgrammer

Reputation: 371

ObjectDetection inference with a SavedModel on TFRecords

I'm trying to perform inference using a SavedModel from the detection model zoo. I was able to get a working example using the Object Detection Tutorial. However, it runs far to slowly and I suspect it could be quicker if I use tfrecords as inputs. Additionally, unfortunately I'm stuck using Tensorflow 1.14. 

    feature_description = {'image_raw': tf.io.FixedLenFeature([], tf.string, default_value='')}
    def _parse_function(example_proto):
        # Parse the input `tf.Example` proto using the dictionary above.
        return tf.io.parse_single_example(example_proto, feature_description)

    sm_dir = '/faster_rcnn_inception_resnet_v2_atrous_coco_2018_01_28/saved_model'
    sm = tf.saved_model.load_v2(sm_dir)
    f = sm.signatures['serving_default']
    tfr_loc = 'inputs.tfr'
    raw_dataset = tf.data.TFRecordDataset(tfr_loc)
    parsed_dataset = raw_dataset.map(_parse_function)

    with tf.Session() as sess:
        for tfr_i in parsed_dataset:
            imgs_raw = tf.decode_raw(tfr_i['image_raw'], tf.float32)
            imgs_raw = tf.reshape(imgs_raw, [-1, 1028, 768, 3])
            inference = f(imgs_raw)
            sess.run(inference)

This gives me the following error:

tensorflow.python.framework.errors_impl.NotFoundError: Resource AnonymousIterator/AnonymousIterator0/N10tensorflow4data16IteratorResourceE does not exist.

Upvotes: 0

Views: 346

Answers (2)

andytaylor823
andytaylor823

Reputation: 51

I know this is an old question, but this came up for me recently, and I found that most of the answers surrounding this issue were unhelpful. I figured I would post what worked for me to potentially help any future troubleshooters.

Via the TF GitHub: https://github.com/tensorflow/tensorflow/issues/33258

A dev suggested passing experimental_run_tf_function=False to model.compile(), and this solved my issue. The output looks odd, but the code runs and my model trains successfully.

Upvotes: 1

Prabhat Kumar Sahu
Prabhat Kumar Sahu

Reputation: 994

You need to use Eager Execution at the beginning of your script.

tf.enable_eager_execution()

Reference: https://www.tensorflow.org/api_docs/python/tf/data/Dataset#iter

Upvotes: 0

Related Questions