I am running a tutorial from the Object Detection API, and I am using the Oxford dataset using ResNet Faster-RCNN.
When I evaluate my trained model by running (eval.py), Tensorboard will return around 0.95 a smoothed precision value.
My question is, how many of the entire set of images does he rate? Because from Tensorboard and their links to the tutorial ( https://github.com/tensorflow/models/blob/master/object_detection/g3doc/running_pets.md ), Tensorboard shows only 10 images.
Does this mean they only check accuracy with 10 images?
My number of centimeters for checking data in Oxford should be about 2200.
In my configuration, I set the input path correctly as follows:
eval_input_reader: {
tf_record_input_reader {
input_path: "my_path/pet_val.record"
}
label_map_path: "my_path/pet_label_map.pbtxt"
shuffle: false
num_readers: 1
}
eval.py mAP ?
eval.py 1 GPU, .
, API F- fps ( )? - ?
edit: , eval , /object _detection/samples/configs/faster_rcnn_resnet101_pets.config#L131. len (result_lists) https://github.com/tensorflow/models/blob/master/object_detection/eval_util.py#L404, 2000, eval num_examples.
fps, .