hszhao
how to set crop size
Hi, I use trian.prototxt add the data layer and loss layer ,when I set crop=347, the error is cudaSuccess out of memory, when I set crop=147, error is number of labels must match number of predictions, could you help me with that? BTW, i also confuse how to make sure source image size is same as label image size?
name: "pspnet101_VOC2012"
layer { name: "data" type: "ImageSegData" top: "data" top: "label" #top: "data_dim" include { phase: TRAIN } transform_param { mirror: true crop_size: 473 mean_value: 104.008 mean_value: 116.669 mean_value: 122.675 scale_factors: 0.5 scale_factors: 0.75 scale_factors: 1 scale_factors: 1.25 scale_factors: 1.5 scale_factors: 1.75 scale_factors: 2.0 } image_data_param { root_folder: "/pspnet/" source: "/pascal_voc_train_aug.txt" batch_size: 2 shuffle: true label_type: PIXEL } }
layer { name: "loss" type: "SoftmaxWithLoss" bottom: "conv6_interp" bottom: "label" top: "loss" include { phase: TRAIN } loss_param { ignore_label: 255 } }
when i make -j8 && make matcaffe, there is one error is MATLAB_DIR must be specified in Makefile.config to build matcaffe. Makefile:511: recipe for target 'matlab/+caffe/private/caffe_.' failed make: *** [matlab/+caffe/private/caffe_.] Error 1
when you make, do you meet this error ?
@landiaokafeiyan I am getting the same error. Did you find the solution for this? I have image size 360X480.
Hi, I met with similar problem. Any idea how you set the crop size without making image size different as label image size?
