any chance you could add a license to this project?

Suggestions:

• MIT
  • about MIT: https://tldrlegal.com/license/mit-license
• Apache v2
  • about Apache v2: https://tldrlegal.com/license/apache-license-2.0-(apache-2.0)

I want to run deeplab.py to show the result image. And I use model 'vgg16_20M.caffemodel ' or 'train2_iter_8000.caffemodel' (download from the DeepLab models)and net 'deploy21.prototxt' in your project.But there is an error: F0705 22:14:09.639734 19135 layer.hpp:378] Check failed: ExactNumTopBlobs() == top.size() (2 vs. 3) MEMORY_DATA Layer produces 2 top blob(s) as output. Did I use the wrong deploy.prototxt or model?

Hi @martinkersner I run deeplab.py using the downloaded official caffemodel(train2_iter_8000.caffemodel). But I get wrong results.The object seems to locate in the wrong position. I have checked the output of each layer and found the output of fully connected layer fc6 incorrect. There are my results.Please give me some advices.Thanks!!

Chen

Hi Martin, I found your tutorial very help and I am following it to train a model. A quick question: do you have code snippet to generate the label image, like this one in your tutorial? Thanks.

https://camo.githubusercontent.com/4df6f507fe1ec20b526af294eeadc482e0f4bf1e/687474703a2f2f692e696d6775722e636f6d2f6e35507a765a552e706e673f31

Hello,

When I want to convert my RGB labels to 1D labels, I get such warnings for each PNG file:

/usr/local/lib/python2.7/dist-packages/skimage/io/_io.py:132: UserWarning: train/0257.png is a low contrast image
  warn('%s is a low contrast image' % fname)

and when I check the PNG files, they are black.

may I ask you what this conversion does actually and the reason of this warning and black PNG labels?

Hello,

I am using python 2.7.When i tried to run mat2png.py,it shows,

KeyError: 'GTcls' and it is right as the function mat2png_hariharan in utils.py,is not assigning the key to the output.I tried may[key]=loadmat....,but still mat[key].Segmentation throws error.Any thoughts will be useful

Hi, minor improvement here, but it might be useful to other people that want to convert the VOC2012 segmentation masks to 1D images; using a vectorized numpy form saves lots of time (10-fold improvement).

I'm following the installation guide, and after I do "make all" I get the following error, which I suspect is from some issue with matio, although it is properly installed and I can access it via dpkg:

[ 88%] Building CXX object tools/CMakeFiles/caffe.bin.dir/caffe.cpp.o Linking CXX executable caffe ../lib/libcaffe.a(blob.cpp.o): In function caffe::Blob<float>::FromMat(char const*)': blob.cpp:(.text._ZN5caffe4BlobIfE7FromMatEPKc[_ZN5caffe4BlobIfE7FromMatEPKc]+0x3d): undefined reference toMat_Open' blob.cpp:(.text._ZN5caffe4BlobIfE7FromMatEPKc[_ZN5caffe4BlobIfE7FromMatEPKc]+0x117): undefined reference to Mat_VarReadInfo' blob.cpp:(.text._ZN5caffe4BlobIfE7FromMatEPKc[_ZN5caffe4BlobIfE7FromMatEPKc]+0x477): undefined reference toMat_VarReadDataLinear' blob.cpp:(.text._ZN5caffe4BlobIfE7FromMatEPKc[_ZN5caffe4BlobIfE7FromMatEPKc]+0x548): undefined reference to Mat_VarFree' blob.cpp:(.text._ZN5caffe4BlobIfE7FromMatEPKc[_ZN5caffe4BlobIfE7FromMatEPKc]+0x55e): undefined reference toMat_VarReadInfo' blob.cpp:(.text._ZN5caffe4BlobIfE7FromMatEPKc[_ZN5caffe4BlobIfE7FromMatEPKc]+0x6b0): undefined reference to Mat_VarReadDataLinear' blob.cpp:(.text._ZN5caffe4BlobIfE7FromMatEPKc[_ZN5caffe4BlobIfE7FromMatEPKc]+0x781): undefined reference toMat_VarFree' blob.cpp:(.text._ZN5caffe4BlobIfE7FromMatEPKc[_ZN5caffe4BlobIfE7FromMatEPKc]+0x790): undefined reference to `Mat_Close'

Have you come across this error, and if so how did you get around it? Thanks.

train.prototxt layers { name: "data" type: IMAGE_SEG_DATA top: "data" top: "label" image_data_param { root_folder: "/media/hao/DATA/DataSet/VOC2012" source: "voc12/list/train.txt" label_type: PIXEL batch_size: 20 shuffle: true } transform_param { mean_value: 104.008 mean_value: 116.669 mean_value: 122.675 crop_size: 306 mirror: true } include: { phase: TRAIN } }

When run " ./run_pascal.sh" for train, it get the error:[libprotobuf ERROR google/protobuf/text_format.cc:288] Error parsing text-format caffe.NetParameter: 21:3: Unknown enumeration value of "IMAGE_SEG_DATA" for field "type". The "IMAGE_SEG_DATA" layer exists in the source code.

I meet this error when I try to add BN layer to this version:

In file included from src/caffe/layers/batch_norm_layer.cpp:4:0:
./include/caffe/layers/batch_norm_layer.hpp: In instantiation of ‘class caffe::BatchNormLayer<float>’:
src/caffe/layers/batch_norm_layer.cpp:249:1:   required from here
./include/caffe/layers/batch_norm_layer.hpp:49:30: error: conflicting return type specified for ‘const char* caffe::BatchNormLayer<Dtype>::type() const [with Dtype = float]’
   virtual inline const char* type() const { return "BatchNorm"; }

I really have no idea how to solve it because I believe it is correct.

Here is the error: make pycaffe find: matlab/+caffe/private': No such file or directory make: *** No rule to make targetinclude/caffe/layers/python_layer.hpp', needed by `python/c affe/_caffe.so'. Stop.

It seems that there is not file in path of `include/caffe/layers/', so how can i download your caffe's version? Thanks!

there is nothing wrong when i make all. but error in make test

LD .build_release/src/caffe/test/test_euclidean_loss_layer.o LD .build_release/src/caffe/test/test_mvn_layer.o .build_release/src/caffe/test/test_contrastive_loss_layer.o: In function caffe::ContrastiveLossLayerTest_TestForward_Test<caffe::CPUDevice<float> >::TestBody()': test_contrastive_loss_layer.cpp:(.text._ZN5caffe41ContrastiveLossLayerTest_TestForward_TestINS_9CPUDeviceIfEEE8TestBodyEv[_ZN5caffe41ContrastiveLossLayerTest_TestForward_TestINS_9CPUDeviceIfEEE8TestBodyEv]+0x307): undefined reference tocaffe::ContrastiveLossParameter_default_instance' .build_release/src/caffe/test/test_contrastive_loss_layer.o: In function caffe::ContrastiveLossLayerTest_TestForward_Test<caffe::GPUDevice<float> >::TestBody()': test_contrastive_loss_layer.cpp:(.text._ZN5caffe41ContrastiveLossLayerTest_TestForward_TestINS_9GPUDeviceIfEEE8TestBodyEv[_ZN5caffe41ContrastiveLossLayerTest_TestForward_TestINS_9GPUDeviceIfEEE8TestBodyEv]+0x307): undefined reference tocaffe::ContrastiveLossParameter_default_instance' .build_release/src/caffe/test/test_contrastive_loss_layer.o: In function caffe::ContrastiveLossLayerTest_TestForwardLegacy_Test<caffe::CPUDevice<float> >::TestBody()': test_contrastive_loss_layer.cpp:(.text._ZN5caffe47ContrastiveLossLayerTest_TestForwardLegacy_TestINS_9CPUDeviceIfEEE8TestBodyEv[_ZN5caffe47ContrastiveLossLayerTest_TestForwardLegacy_TestINS_9CPUDeviceIfEEE8TestBodyEv]+0x336): undefined reference tocaffe::ContrastiveLossParameter_default_instance' .build_release/src/caffe/test/test_contrastive_loss_layer.o: In function caffe::ContrastiveLossLayerTest_TestForwardLegacy_Test<caffe::GPUDevice<float> >::TestBody()': test_contrastive_loss_layer.cpp:(.text._ZN5caffe47ContrastiveLossLayerTest_TestForwardLegacy_TestINS_9GPUDeviceIfEEE8TestBodyEv[_ZN5caffe47ContrastiveLossLayerTest_TestForwardLegacy_TestINS_9GPUDeviceIfEEE8TestBodyEv]+0x336): undefined reference tocaffe::ContrastiveLossParameter_default_instance' .build_release/src/caffe/test/test_contrastive_loss_layer.o: In function caffe::ContrastiveLossLayerTest_TestForward_Test<caffe::CPUDevice<double> >::TestBody()': test_contrastive_loss_layer.cpp:(.text._ZN5caffe41ContrastiveLossLayerTest_TestForward_TestINS_9CPUDeviceIdEEE8TestBodyEv[_ZN5caffe41ContrastiveLossLayerTest_TestForward_TestINS_9CPUDeviceIdEEE8TestBodyEv]+0x304): undefined reference tocaffe::ContrastiveLossParameter_default_instance' .build_release/src/caffe/test/test_contrastive_loss_layer.o:test_contrastive_loss_layer.cpp:(.text._ZN5caffe41ContrastiveLossLayerTest_TestForward_TestINS_9GPUDeviceIdEEE8TestBodyEv[_ZN5caffe41ContrastiveLossLayerTest_TestForward_TestINS_9GPUDeviceIdEEE8TestBodyEv]+0x304): more undefined references to caffe::_ContrastiveLossParameter_default_instance_' follow collect2: error: ld returned 1 exit status Makefile:620: recipe for target '.build_release/test/test_contrastive_loss_layer.testbin' failed make: *** [.build_release/test/test_contrastive_loss_layer.testbin] Error 1 make: *** Waiting for unfinished jobs.... .build_release/src/caffe/test/test_pooling_layer.o: In functioncaffe::PoolingLayer::MaxTopBlobs() const': test_pooling_layer.cpp:(.text._ZNK5caffe12PoolingLayerIfE11MaxTopBlobsEv[_ZNK5caffe12PoolingLayerIfE11MaxTopBlobsEv]+0x23): undefined reference to caffe::_PoolingParameter_default_instance_' .build_release/src/caffe/test/test_pooling_layer.o: In functioncaffe::PoolingLayer::MaxTopBlobs() const': test_pooling_layer.cpp:(.text._ZNK5caffe12PoolingLayerIdE11MaxTopBlobsEv[_ZNK5caffe12PoolingLayerIdE11MaxTopBlobsEv]+0x23): undefined reference to caffe::_PoolingParameter_default_instance_' collect2: error: ld returned 1 exit status Makefile:620: recipe for target '.build_release/test/test_pooling_layer.testbin' failed make: *** [.build_release/test/test_pooling_layer.testbin] Error 1 .build_release/src/caffe/test/test_rnn_layer.o: In functioncaffe::RecurrentLayer::ExactNumTopBlobs() const': test_rnn_layer.cpp:(.text._ZNK5caffe14RecurrentLayerIfE16ExactNumTopBlobsEv[_ZNK5caffe14RecurrentLayerIfE16ExactNumTopBlobsEv]+0xdb): undefined reference to caffe::_RecurrentParameter_default_instance_' .build_release/src/caffe/test/test_rnn_layer.o: In functioncaffe::RecurrentLayer::MinBottomBlobs() const': test_rnn_layer.cpp:(.text._ZNK5caffe14RecurrentLayerIfE14MinBottomBlobsEv[_ZNK5caffe14RecurrentLayerIfE14MinBottomBlobsEv]+0xdb): undefined reference to caffe::_RecurrentParameter_default_instance_' .build_release/src/caffe/test/test_rnn_layer.o: In functioncaffe::RecurrentLayer::ExactNumTopBlobs() const': test_rnn_layer.cpp:(.text._ZNK5caffe14RecurrentLayerIdE16ExactNumTopBlobsEv[_ZNK5caffe14RecurrentLayerIdE16ExactNumTopBlobsEv]+0xdb): undefined reference to caffe::_RecurrentParameter_default_instance_' .build_release/src/caffe/test/test_rnn_layer.o: In functioncaffe::RecurrentLayer::MinBottomBlobs() const': test_rnn_layer.cpp:(.text._ZNK5caffe14RecurrentLayerIdE14MinBottomBlobsEv[_ZNK5caffe14RecurrentLayerIdE14MinBottomBlobsEv]+0xdb): undefined reference to caffe::_RecurrentParameter_default_instance_' .build_release/src/caffe/test/test_rnn_layer.o: In functioncaffe::RecurrentLayer::MaxBottomBlobs() const': test_rnn_layer.cpp:(.text._ZNK5caffe14RecurrentLayerIdE14MaxBottomBlobsEv[_ZNK5caffe14RecurrentLayerIdE14MaxBottomBlobsEv]+0xf3): undefined reference to caffe::_RecurrentParameter_default_instance_' .build_release/src/caffe/test/test_rnn_layer.o:test_rnn_layer.cpp:(.text._ZNK5caffe14RecurrentLayerIfE14MaxBottomBlobsEv[_ZNK5caffe14RecurrentLayerIfE14MaxBottomBlobsEv]+0xf3): more undefined references tocaffe::RecurrentParameter_default_instance' follow collect2: error: ld returned 1 exit status Makefile:620: recipe for target '.build_release/test/test_rnn_layer.testbin' failed make: *** [.build_release/test/test_rnn_layer.testbin] Error 1

looking forward your help

In VGG, the first connected layer is 4096 7x7 filters, how to adapt these param to 3x3 1024 or 4x4 4096? Is it just copy parameters in order and remove the others? Thanks a lot.

I tested resnet-101 successfully, but it's not the same as the author's miou. The result before densecrf on the validation set was 0.7646 and the paper was 0.7635. Strangely, the miou I ran out of VGG-16 was consistent with the paper. Does anyone have the same problem?

I use deeplab to do segmentation.when I use deeplab.py I meet this error:

Traceback (most recent call last):
  File "deeplab.py", line 106, in <module>
    main()
  File "deeplab.py", line 24, in main
    net = Segmenter(net_path, model_path, gpu_id)
  File "/home/muses/train-DeepLab/segmenter.py", line 17, in __init__
    caffe.Net.__init__(self, prototxt, model)
Boost.Python.ArgumentError: Python argument types in
    Net.__init__(Segmenter, str, str)
did not match C++ signature:
    __init__(boost::python::api::object, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >, int)
    __init__(boost::python::api::object, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > network_file, int phase, int level=0, boost::python::api::object stages=None, boost::python::api::object weights=None)

my command is:

python deeplab.py 0 ./exper/0822/config/PSP101-ele/pspnet101_ele_8s.prototxt ./exper/0822/model/model_iter_10000.caffemodel ./test/pic/87.png

how to solve this problem?

Hi, I am try to reproduce the deeplab results using VOC2012 data set as described in instructions at https://bitbucket.org/aquariusjay/deeplab-public-ver2 I tried two approaches 1) using already trained model vgg16_20M.caffemodel and created a symoblic link of this model with name init.caffemodel in deeplab_largeFOV directory and fine tuned it secondly I tried to train without using init.caffemode. In first case training is going fine and prediction accuracies are good. But when i use second approach, i observe that only background class is predicted and test, evaluation phases show that only background class has some accuracy whereas all other classes are predicted to be zero. Can you please guide in this context. Best