Failed to load finetune_checkpoint in object detection training
I am following this tutorial .
I am using Oxford-IIIT Pet data and config file similar to this .
In the pipe_config file, I have specified the exact breakpoint as
fine_tune_checkpoint: "{PATH TO}/ssd_mobilenet_v1_coco_11_06_2017/model.ckpt"
from_detection_checkpoint: false
But when i run python object_detection / train.py ... --train_dir = $ {PATH_TO_TRAIN_DIR}
I am getting a list of warnings.
WARNING: root: The [MobilenetV1 / Conv2d_0 / BatchNorm / beta] variable is not available at the checkpoint.
...
...
...
WARNING: root: variable [MobilenetV1 / Conv2d_9_pointwise / weight] not available at checkpoint
and program exits saying
ValueError: no variables to save
I have checked ckpt in python using
print_tensors_in_checkpoint_file('{PATH TO}/ssd_mobilenet_v1_coco_11_06_2017/model.ckpt/model.ckpt', '',"")
The output was a list.
FeatureExtractor / MobilenetV1 / Conv2d_0 / BatchNorm / beta (DT_FLOAT) [32]
...
...
...
FeatureExtractor / MobilenetV1 / Conv2d_9_pointwise / weight (DT_FLOAT) [1,1,512,512]
Note : the training script runs fine when finetune_checkpoint is disabled
from_detection_checkpoint: true
in the pipeline config file
source to share
Tuning is from_detection_checkpoint: true
really necessary for the checkpoint file you are using because this model has been trained as a detector (so nothing is disabled).
You would only install from_detection_checkpoint: false
if you wanted to train, starting with one of the Slim models , which was pre-prepared for the Imagenet classification.
source to share