I noticed that there are some big changes in the last update of the object detection API in the last days. Maybe it is a good idea to use some tags which are related to the update summaries in the ReadMe of the object detection folder. With tags it would be much easier to switch between different versions.
Yeah I agree, these changes totally broke my workflow...nothing works anymore and I don't know how to get things back to working state...
I do not know what happened so much confusion recently, I almost collapsed
Agreed. when I get through one error, another pops up.
+1
Also could you add TF version (1.3, 1.4rc, nightly) everytime?
Does anyone know what tensorflow version is working for the code in models/research/object_detection?
Guys Take a look at this REPO.. They have implemented the MASK RCNN
https://github.com/matterport/Mask_RCNN
and its perfect. Hope it helps
I found this branch still works with TensorFlow 1.3
git checkout 9cfd2d93062bb7a44c608fa46d1c55cc297dfe76
If you want to know, this is the merge that breaks everything. I commented on it but no replies.
https://github.com/tensorflow/models/pull/2639
Apologies for the breaking changes. Most issues should be fixed now. This is a great idea. In the future we will tag major changes with a compatible TF version.
Please feel free to report problems as you see them. I will close this issue now and we can handle specific problems in separate issues.
Most helpful comment
Yeah I agree, these changes totally broke my workflow...nothing works anymore and I don't know how to get things back to working state...