Where are "img0_b", "img1_b", and "flow_gt_b" defined in FlowNet2 model?

[brjeon]

I have encountered "Unknown bottom blob 'img0_b' (layer 'Concat1', bottom index 1)" error when training with FlowNet2 model.

The follwing is the part of "train.prototxt" which is originally "FlowNet2_train.prototxt.template".

layer { name: "Concat1" type: "Concat" bottom: "img0_a" bottom: "img0_b" top: "img0" concat_param { axis: 0 } } layer { name: "Concat2" type: "Concat" bottom: "img1_a" bottom: "img1_b" top: "img1" concat_param { axis: 0 } } layer { name: "Concat3" type: "Concat" bottom: "flow_gt_a" bottom: "flow_gt_b" top: "flow_gt" concat_param { axis: 0 } }

Thank you in advance.

[linrio]

I meet the same problem and I guess it myself Another question: ResampleLayer cannot do backward.

[eddy-ilg]

This is actually because when finetuning for small displacements we use a mixture of small and large displacements, which we get from different databases. Therefore there is img0_a and img0_b (later the batches are concatenated). However, I will fix the model so it works with only one dataset.

[CLT29]

@eddy-ilg I am getting the same issue - "ResampleLayer cannot do backward" did you figure it out?

[tachim]

Hi @eddy-ilg, any word on your fix?

[dakeli]

LayerFactoryTest/0.TestCreateLayer src/caffe/test/test_layer_factory.cpp:47: Failure Value of: layer->type() Actual: "NormLayer" Expected: iter->first Which is: "ChannelNorm" F0827 18:42:06.915846 28432 custom_data_layer.cpp:661] Check failed: !pthreadjoin(thread, NULL) Pthread joining failed. Check failure stack trace: @ 0x2b003d54bdaa (unknown) @ 0x2b003d54bce4 (unknown) @ 0x2b003d54b6e6 (unknown) @ 0x2b003d54e687 (unknown) @ 0x2b003f858132 caffe::CustomDataLayer<>::JoinPrefetchThread() @ 0x2b003f8598e4 caffe::CustomDataLayer<>::~CustomDataLayer() @ 0x2b003f859bd9 caffe::CustomDataLayer<>::~CustomDataLayer() @ 0x71821b caffe::LayerFactoryTest_TestCreateLayer_Test<>::TestBody() @ 0x8f48b3 testing::internal::HandleExceptionsInMethodIfSupported<>() @ 0x8eb597 testing::Test::Run() @ 0x8eb63e testing::TestInfo::Run() @ 0x8eb745 testing::TestCase::Run() @ 0x8eea88 testing::internal::UnitTestImpl::RunAllTests() @ 0x8eed17 testing::UnitTest::Run() @ 0x46d2df main @ 0x2b0040b04830 (unknown) @ 0x474e99 _start @ (nil) (unknown) make: *** [runtest] 已放弃 (core dumped)

anyone can help me？

[nikolausmayer]

@dakeli are you running a test mode, e.g. make runtest? Our own custom layers do not run tests and will fail.

[shahabty]

Hello, @eddy-ilg Unknown bottom blob 'img0_a_org' (layer 'crop_params', bottom index 0) Is there any solution?

[QM117]

@shahabty I met the same question. It's because that when I was just running the FlowNet2 model, I typed FlowNet2_train.prototxt.template in the terminal rather than FlowNet2_deploy.prototxt.template. That is my typing mistake. I hope it can help you.

[cdancette]

I'm facing the same issue, any info about this ?

[dudeDoff]

Same issue here. The final FlowNet2 model still has layers referencing to img0_a, img1_b and flow_gt_b. Furthermore, the two CustomDataLayer at the very top have unconnected top blobs: blob0-2. Is there a solution?

[TusharNimbhorkar]

@nikolausmayer Kind of getting the same error when I test on Flownet-2 Any idea why this might be happening?

[OPzealot]

@TusharNimbhorkar Same error here. Have you solved it?

[vivasvan1]

same problem. Any updates?