A: For the object detecion framework to be TPU-complient, we must pad our input
tensors to static shapes. This means that we must pad to a fixed number of
bounding boxes, configured by InputReader.max_number_of_boxes
. It is
important to set this value to a number larger than the maximum number of
groundtruth boxes in the dataset. If an image is encountered with more
bounding boxes, the excess boxes will be clipped.
A: This BackupHandler (tf_slim.tfexample_decoder.BackupHandler) was introduced in tensorflow 1.5.0 so runing with earlier versions may cause this issue. It now has been replaced by object_detection.data_decoders.tf_example_decoder.BackupHandler. Whoever sees this issue should be able to resolve it by syncing your fork to HEAD. Same for LookupTensor.
A: Similar to BackupHandler, syncing your fork to HEAD should make it work.
A: The inference time reported in model zoo is mean time of testing hundreds of images with an internal machine. As mentioned in TensorFlow detection model zoo, this speed depends highly on one's specific hardware configuration and should be treated more as relative timing.