Models: Object detection API, training part: NewRandomAccessFile failed to Create/Open: data/object-detection.pbtxt

[Gerahal]

System information

• What is the top-level directory of the model you are using: Tensorflow\models\research\object_detection
• Have I written custom code (as opposed to using a stock example script provided in TensorFlow): No, followed the tutorial about Object Detection API made by sentdex
• OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Windows 10
• TensorFlow installed from (source or binary): pip install, GPU version of Tensorflow
• TensorFlow version (use command below): 1.13.1
• Bazel version (if compiling from source): N/A
• CUDA/cuDNN version: 10.0.130 / 7.3.1
• GPU model and memory: NVIDIA GTX970M, 3GB
• Exact command to reproduce: python legacy\train.py --logtostderr --pipeline_config_path=trainingfaster_rcnn_inception_v2_pets.config --train_dir=training\

Describe the problem

When I try this command in the anaconda prompt:
(from ...Tensorflow\models\research\object_detection)

python legacy/train.py
--train_dir=training/
--pipeline_config_path=training/faster_rcnn_inception_v2_pets.config
--logtostderr

It should start training to detect the objects I want but I get this error:

tensorflow.python.framework.errors_impl.NotFoundError: NewRandomAccessFile failed to Create/Open: data/object-detection.pbtxt : Le fichier sp\udce9cifi\udce9 est introuvable.
; No such file or directory

And I can't figure why I get this error, the object-detection.pbtxt is in the right place, I have also tried to put entire path ( C:/folder1/folder2/.../data/object-detection.pbtxt ) in the config file. I tried to use normal / double slash or backslash in my command line and inside the config file.

Is it because of object-detection.pbtxt that is considered as a .txt file when I created it with the notepad ?
Hoping to find a solution quick.

Source code / logs

Here is the part of the CONFIG file where I specified the different paths:

train_input_reader: {
  tf_record_input_reader {
    input_path: "data\train.record-?????-of-00010"
  }
  label_map_path: "data\objectdetection.pbtxt"
}

eval_config: {
  metrics_set: "coco_detection_metrics"
  num_examples: 1101
}

eval_input_reader: {
  tf_record_input_reader {
    input_path: "data\test.record-?????-of-00010"
  }
  label_map_path: "data\objectdetection.pbtxt"
  shuffle: false
  num_readers: 1
}

[galadash]

For me the combination of escaping the backslash AND using the full path worked:

eval_input_reader: {
  tf_record_input_reader {
    input_path:  "D:\\Projects\\TFruns\\Run1\\data\\eval.record"
  }
  label_map_path: "D:\\Projects\\TFruns\\Run1\\data\\labelmap.pbtxt"
  shuffle: false
  num_readers: 1
}

Hope this helps.

In any case, I'm calling model_main: python object_detection/model_main.py, which might handle things differently.. Not sure, but you could try to use that instead.

[Arunpoochelvan]

For me the combination of escaping the backslash AND using the full path worked:

eval_input_reader: {
  tf_record_input_reader {
    input_path:  "D:\\Projects\\TFruns\\Run1\\data\\eval.record"
  }
  label_map_path: "D:\\Projects\\TFruns\\Run1\\data\\labelmap.pbtxt"
  shuffle: false
  num_readers: 1
}

Hope this helps.

In any case, I'm calling model_main: python object_detection/model_main.py, which might handle things differently.. Not sure, but you could try to use that instead.

Thank you bro you solved my day