objection detection data augmentation - random_jpeq_quality error

[ecatkins]

System information

• What is the top-level directory of the model you are using: research/object_detection
• Have I written custom code (as opposed to using a stock example script provided in TensorFlow): no
• OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Debian GNU/Linux 9 (stretch)
• TensorFlow installed from (source or binary): not sure (preinstalled GCP Deep Learning VM)
• TensorFlow version (use command below): 1.131
• Bazel version (if compiling from source): 0.15.2
• CUDA/cuDNN version: 10/7
• GPU model and memory: Tesla P4 8GB
• Exact command to reproduce: python objection_detection/model_main.py

Describe the problem

I have tried to add in different data augumentations to the config file to train a Faster R-CNN with Resnet-101. When I try to add random_jpeg_quality I receive the following error

  File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/ops/gen_image_ops.py", line 1324, in encode_jpeg
    quality = _execute.make_int(quality, "quality")
  File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/eager/execute.py", line 108, in make_int
    (arg_name, repr(v)))
TypeError: Expected int for argument 'quality' not <tf.Tensor 'RandomJpegQuality/cond/random_uniform:0' shape=() dtype=int32>.

This is the way I have added it to the data_augmentation_options in the config file (but I have also tried it with no arguments i.e. just leaving it to the defaults)

    random_jpeg_quality {
      random_coef: 0.5
      min_jpeg_quality: 40
      max_jpeg_quality: 90
    }

[roger1993]

I met the same problem, random_jpeg_quality doesnt work.