Tensorflow object detection training makes python crash

Question

I am training ssd_mobilenet_v1_coco_2017_11_17 model from Tensorflow object detection model zoo. My dataset is satellite imagery and my aim is to detect vehicles in the images. But training fails with python memory issues. I am training on CPU and my windows 10 machine has 32 gb RAM. TF record file for training is aroung 1.7 GB in size.

I am unable to detect the reason for this failure.

Please help.

Answer 1

try to adjust batch size, queue capacity, reader threads.