"raise ValueError('First step cannot be zero.')" in the training stage about ourcamera

[danp]

It looks like object detection was recently changed to add this error in tensorflow/models#3656. Maybe try using the commit before that was merged, such as with:

git clone https://github.com/tensorflow/models
cd models
git checkout -b before-object-detection-changes 2913cb24ecb16ba955006b41072adce45c5a0f62

[Bellspringsteen]

Hello,

Great you created the .records. You make the train and eval empty folders yourself. Then when you call train.py it will fill those folders with checkpoints and frozen points for models that you can use later.

[danp]

Looking at tensorflow/models#3794, maybe this is no longer valid?

OurCamera/data/faster_rcnn_resnet101_cars.config

Line 92 in 796dbfb

step: 0