Eval.py just use only one image to evaluate, don't use other images

[cannguyen275]

System information

• What is the top-level directory of the model you are using: models/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): Linux Ubuntu 17.04
• TensorFlow installed from (source or binary): conda install
• TensorFlow version (use command below): 1.7.0
• Bazel version (if compiling from source):
• CUDA/cuDNN version: CUDA® Toolkit 9.0; cuDNN v7.1
• GPU model and memory: GForce GTX 1080Ti (11GB)
• Exact command to reproduce:

  python train.py --logtostderr --train_dir=training_point/ --pipeline_config_path=training/faster_rcnn_inception_resnet_v2_atrous_coco.config

  python eval.py --logtostderr --pipeline_config_path=training/faster_rcnn_inception_resnet_v2_atrous_coco.config --checkpoint_dir=training_point/ --eval_dir=eval/

  tensorboard --logdir=training:training_point/,testing:eval/

  Describe the problem

  I'm training a model to detect some object. But when i run eval.py, it just uses only one image to evaluate, even I have creating more than 200 images for evaluate. Following image show the image that tensorflow use for evaluate. You can see it just uses only one image This image shows my tensorboard, only "hand watch" category have the value I don't know why it happens, pls help. It's my pleasure to have your helps! Thanks so much!

Source code / logs

My evaluate config:

eval_config: {
  num_examples: 212
  # Note: The below line limits the evaluation process to 10 evaluations.
  # Remove the below line to evaluate indefinitely.
  num_visualizations: 20
}

eval_input_reader: {
  tf_record_input_reader {
    input_path: "/home/thangtran/ThangTran/Can/rpa/object_detection/test.record"
  }
  label_map_path: "/home/thangtran/ThangTran/Can/rpa/object_detection/training/labelmap.pbtxt"
  shuffle: true
}

[pkulzc]

Did you sync your fork to latest? It might be relevant to this https://github.com/tensorflow/models/pull/4205/files.

[pkulzc]

Is this still happening?

[ymodak]

Closing due to lack of recent activity. Please update the issue when new information becomes available, and we will reopen the issue. Thanks!

[ibrahim1611]

Hi I am testing 26 images on Faster RCNN model using my own dataset. When I am running eval.py it only shows 10 images and dot in the middle of each graph on Tensorboard Please guide.