when i run this demo, there is a error Check failed: error == cudaSuccess (2 vs. 0) out of memory.
my gpu is k80, i think the memory is enough to run this demo ,do you know why ?

i do not know what is the content of test_feat_conv34atrous_v2.prototxt which is a binary type and I do not know how to generate it.

when I run faster_rcnn_build, I got the erro
The destination directory "bin" cannot be found.

Error in faster_rcnn_build (line 13)
mex -O -outdir bin ...

I am very sorry to disturb you !
I cannot find the solution

when I run script_rpn_pedestrian_VGG16_caltech.m
the error comes as below:

Warrning: no windows proposal is loaded !
Reference to non-existent field 'sizes'.

Error in roidb_from_caltech (line 52)
height = imdb.sizes(1,1);

Error in Dataset.caltech_test (line 9)
dataset.roidb_test = dataset.imdb_test.roidb_func(dataset.imdb_test, false);

Error in script_rpn_pedestrian_VGG16_caltech (line 32)
dataset = Dataset.caltech_test(dataset, 'test');

I am very appreciate you to answer this!!!!

I am curious why my MATLAB always crashed when I run the demo the second time.

Hi,

Thanks for sharing your code.

I've downloaded VGG16_caltech_final.zip, and built the code. When I run the demo, there are no pedestrians detected. The scores produced by the adaboost classifier are all very low (essentially zero). The region proposal values and scores seem reasonable. When I examine the "feat" vectors, most of the values are exactly zero. I'm guessing this is the problem. Any suggestions for debugging this would be very welcome.

My setup: Ubuntu 14.04, Cuda 8.0, CuDNN version 2, Nvidia GTX 1080, Matlab R2016b

Many thanks

after run "make all", the error looks like this:

CXX src/caffe/syncedmem.cpp
In file included from ./include/caffe/blob.hpp:9:0,
from src/caffe/blob.cpp:4:
./include/caffe/proto/caffe.pb.h:17:2: error: #error This file was generated by an older version of protoc which is
#error This file was generated by an older version of protoc which is
^
./include/caffe/proto/caffe.pb.h:18:2: error: #error incompatible with your Protocol Buffer headers. Please
#error incompatible with your Protocol Buffer headers. Please
^
./include/caffe/proto/caffe.pb.h:19:2: error: #error regenerate this file with a newer version of protoc.
#error regenerate this file with a newer version of protoc.

Then I simply run "pip install protobuf" and the error is gone, but when I run "make test", there is a new error reads like this:

CXX src/gtest/gtest-all.cpp
src/caffe/test/test_im2col_kernel.cu(88): error: no instance of function template "caffe::im2col_cpu" matches the argument list
argument types are: (const float *, int, int, int, int, int, int, int, int, int, float *)
detected during:
instantiation of "void caffe::Im2colKernelTest_TestGPU_Test<gtest_TypeParam_>::TestBody() [with gtest_TypeParam_=float]"
./src/gtest/gtest.h(7209): here
......

2 errors detected in the compilation of "/tmp/tmpxft_00004d75_00000000-16_test_im2col_kernel.compute_50.cpp1.ii".
make: *** [.build_release/cuda/src/caffe/test/test_im2col_kernel.o] 错误 1

Could you please give some advice? Thanks a lot!
@zhangliliang

Hi,
Thank you for your sharing~ I just follow the install guidance to setup this one,and i have used the caffe you recommend and make matcaffe,then move this mex. to the path you just told me, but when i run the demo i still encountered the follow error:
Warning: You are using gcc version '5.4.0-6ubuntu1~16.04.4)'. The version of gcc is not supported. The version currently supported with MEX is '4.7.x'. For a list of currently supported compilers see: http://www.mathworks.com/support/compilers/current_release.
MEX completed successfully.

can you give me some advise? Thx ~

[libprotobuf ERROR google/protobuf/text_format.cc:245] Error parsing text-format caffe.NetParameter: 534:30: Message type "caffe.ConvolutionParameter" has no field named "filter_stride".
WARNING: Logging before InitGoogleLogging() is written to STDERR
F1019 18:58:15.951474 2862 upgrade_proto.cpp:928] Check failed: ReadProtoFromTextFile(param_file, param) Failed to parse NetParameter file: /home/handson/RPN_BF/output/VGG16_caltech_final/rpn_test.prototxt
*** Check failure stack trace: ***
已杀死
handson@handson-Aspire-TC-705:~$

I dont use GPU.Could you please give some advice? Thanks a lot!

I followed the instruction in "Testing Demo" to see the detection results on my own dataset, but I got very low recall disagree with the results in the paper. Did I get something wrong? Thanks a lot.
Here is the ROC I got:

i can not find faster_rcnn_build

the error reads like this:

Caught "std::exception" Exception message is:
CHECK failed: generated_database_->Add(encoded_file_descriptor, size):

Thanks very much! @zhangliliang

I ran the code and found that some of the regression losses aren't consistent. But the final detection result is good. I wonder what's the problem

Hello, I tried to compile RPN_BF customized caffe but got an cudnn ralated error.

The error message is given below:
PROTOC src/caffe/proto/caffe.proto
CXX .build_release/src/caffe/proto/caffe.pb.cc
CXX src/caffe/internal_thread.cpp
In file included from ./include/caffe/util/device_alternate.hpp:40:0,
from ./include/caffe/common.hpp:19,
from ./include/caffe/internal_thread.hpp:4,
from src/caffe/internal_thread.cpp:2:
./include/caffe/util/cudnn.hpp: In function ‘void caffe::cudnn::createPoolingDesc(cudnnPoolingStruct*, caffe::PoolingParameter_PoolMethod, cudnnPoolingMode_t, int, int, int, int, int, int)’:
./include/caffe/util/cudnn.hpp:123:3: error: too few arguments to function ‘cudnnStatus_t cudnnSetPooling2dDescriptor(cudnnPoolingDescriptor_t, cudnnPoolingMode_t, cudnnNanPropagation_t, int, int, int, int, int, int)’
In file included from ./include/caffe/util/cudnn.hpp:5:0,
from ./include/caffe/util/device_alternate.hpp:40,
from ./include/caffe/common.hpp:19,
from ./include/caffe/internal_thread.hpp:4,
from src/caffe/internal_thread.cpp:2:
/usr/local/include/cudnn.h:803:27: note: declared here
Makefile:521: recipe for target '.build_release/src/caffe/internal_thread.o' failed
make: *** [.build_release/src/caffe/internal_thread.o] Error 1

My environment is ubuntu 15.04, with cuda 5.7 cudnn v5. I suspect the customized caffe is outdated and does not support cudnn v5. Any suggestions will be appreciated!

Hi,

I ran script_rpn_pedestrian_VGG16_caltech and received the error message as following.

`Loading region proposals...done
Warrning: no windows proposal is loaded !
Reference to non-existent field 'sizes'.

Error in roidb_from_caltech (line 52)
height = imdb.sizes(1,1);

Error in Dataset.caltech_trainval>@(x)x.roidb_func(x,false) (line 6)
dataset.roidb_train = cellfun(@(x) x.roidb_func(x, false),
dataset.imdb_train, 'UniformOutput', false);

Error in Dataset.caltech_trainval (line 6)
dataset.roidb_train = cellfun(@(x) x.roidb_func(x, false),
dataset.imdb_train, 'UniformOutput', false);

Error in script_rpn_pedestrian_VGG16_caltech (line 31)
dataset = Dataset.caltech_trainval(dataset, 'train');

Error in run (line 96)
evalin('caller', [script ';']);`

The demo script (script_rpn_bf_pedestrian_VGG16_caltech_demo) works fine, and I already run extract_img_anno.m to achieve train and test folder. Kindly give suggestion.

RPN_BF startup done
GPU 1: free memory 11421024256
GPU 2: free memory 7635664896
GPU 3: free memory 7606304768
Use GPU 1
Cleared 0 solvers and 1 stand-alone nets

stage one RPN

Doing nms ... nms: 77 / 1155
nms: 154 / 1155
nms: 230 / 1155
nms: 317 / 1155
nms: 395 / 1155
nms: 480 / 1155
nms: 568 / 1155
nms: 651 / 1155
nms: 724 / 1155
nms: 802 / 1155
nms: 874 / 1155
nms: 949 / 1155
nms: 1035 / 1155
nms: 1121 / 1155
aver_boxes_num = 965, select top 40
gt recall rate = 1.0000
Preparing the results for Caltech evaluation ...Done.Loading detections: /home/dgc/rpnbf/external/code3.2.1/results/UsaTest
Loading ground truth: /home/dgc/rpnbf/external/code3.2.1/results/UsaTest
Experiment #1: Reasonable

gts =

{1x4024 cell}

dts =

Empty cell array: 1-by-0

Evaluating: /home/dgc/rpnbf/external/code3.2.1/results/UsaTest
Index exceeds matrix dimensions.

Error in dbEval_RPNBF>plotExps (line 193)
stre={res(:,1).stre};

Error in dbEval_RPNBF (line 127)
plotExps( res, plotRoc, plotAlg, plotNum, plotName, ...

Error in Faster_RCNN_Train.do_proposal_test_caltech (line 62)
dbEval_RPNBF;

Error in script_rpn_pedestrian_VGG16_caltech (line 53)
Faster_RCNN_Train.do_proposal_test_caltech(conf_proposal, model.stage1_rpn, dataset.imdb_test, dataset.roidb_test, cache_name, method_name);

when I run extract_img_anno it has the error blow,could you tell me the reason,I cannot figure it out!!
Thank you for your answer!!

extract_img_anno
Index exceeds matrix dimensions.

Error in seqReaderPlugin>getTs (line 195)
fseek(fid,info.seek(frame+1),'bof');

Error in seqReaderPlugin>open (line 133)
ts = getTs( 0:(n-1), fid, info );

Error in seqReaderPlugin (line 47)
[infos{h},fids(h),tNms{h}]=open(fName,info); return;

Error in seqIo>reader (line 113)
r=@seqReaderPlugin; s=r('open',int32(-1),fName);

Error in seqIo (line 64)
case {'reader','r'}, out = reader( fName, varargin{:} );

Error in dbExtract (line 40)
sr=seqIo([pth '/videos/' name '.seq'],'reader'); info=sr.getinfo();

Error in extract_img_anno (line 16)
dbExtract([dataDir type],1,skip);

Hi, I want to know why you add these two parameter?

I have modified the demo code to generate the bounding boxes and confidence scores on Caltech dataset, but however, after setting the same hyper-parameters (including NMS thresholds and some others) as those in the training code, we cannot actually reproduce the reported MR value (as reported in ECCV paper, which is about 10%). After delving into the details, the bounding boxes generated by RPN are of high recall rate (>= 95%) and gives fairly well confidence scores (MR ~ 16%). But after using the BF (adaboost), the proposals are mostly rescored with negative values, and perform very poorly on the test data (MR >= 40%), so we are wondering if there might be some differences between the demo code and the training code (apart from those hyper-parameters).

B.T.W, if we directly apply the pre-trained model on the ETH and INRIA dataset, the generated proposals will be suffered from a very poor MR value (also with a low recall rate). Is this mainly because the models are not finetuned on INRIA? If so, we will be grateful if you can provide the trained weights on the INRIA.

Results on Caltech dataset:
{'test': {'recall': 98.76543209876543, 'mr': 16.040030572979582, 'proposal number per image': 91.115556660039758, 'precision': 0.30723820287596904, 'ap': 81.441812506102707}}

function script_rpn_bf_pedestrian_VGG16_caltech_demo()
close all;
clc;
clear mex;
clear is_valid_handle; % to clear init_key
run(fullfile(fileparts(fileparts(mfilename('fullpath'))), 'startup'));
%% -------------------- CONFIG --------------------
opts.caffe_version          = 'caffe_faster_rcnn';
opts.gpu_id		    = 5;
active_caffe_mex(opts.gpu_id, opts.caffe_version);

opts.use_gpu                = true;

opts.test_scales            = 720;
opts.test_max_size          = 960;
opts.feat_stride            = 16;
opts.test_binary            = false;
opts.test_min_box_size      = 16;
opts.test_min_box_height    = 50;
opts.test_drop_boxes_runoff_image = true;


%% -------------------- INIT_MODEL --------------------
model_dir                   = fullfile(pwd, 'output', 'VGG16_caltech_final'); 


rpn_bf_model.rpn_net_def = fullfile(model_dir, 'rpn_test.prototxt');
rpn_bf_model.rpn_net = fullfile(model_dir, 'final');
rpn_bf_model.bf_net_def = fullfile(model_dir, 'bf_test.prototxt');

rpn_bf_model.conf_rpn.test_scales = opts.test_scales;
rpn_bf_model.conf_rpn.test_max_size = opts.test_max_size;
rpn_bf_model.conf_rpn.max_size = opts.test_max_size;
rpn_bf_model.conf_rpn.feat_stride = opts.feat_stride;
rpn_bf_model.conf_rpn.test_binary = opts.test_binary;
rpn_bf_model.conf_rpn.test_min_box_size = opts.test_min_box_size;
rpn_bf_model.conf_rpn.test_min_box_height = opts.test_min_box_height;
rpn_bf_model.conf_rpn.test_drop_boxes_runoff_image = opts.test_drop_boxes_runoff_image;

rpn_bf_model.conf_bf.test_scales = opts.test_scales;
rpn_bf_model.conf_bf.test_max_size = opts.test_max_size;
rpn_bf_model.conf_bf.max_size = opts.test_max_size;

ld = load(fullfile(model_dir, 'mean_image'));  
rpn_bf_model.conf_rpn.image_means = gpuArray(ld.image_mean);
rpn_bf_model.conf_bf.image_means = gpuArray(ld.image_mean);
clear ld;

ld = load(fullfile(model_dir, 'anchors'));  
rpn_bf_model.conf_rpn.anchors = ld.anchors;
clear ld;

rpn_net = caffe.Net(rpn_bf_model.rpn_net_def, 'test');
rpn_net.copy_from(rpn_bf_model.rpn_net);

fast_rcnn_net = caffe.Net(rpn_bf_model.bf_net_def, 'test');

BF_prototxt_path = fullfile('models', 'test_feat_conv34atrous_v2.prototxt');

caffe_net = caffe.Net(BF_prototxt_path, 'test');
caffe_net.copy_from(rpn_bf_model.rpn_net);
caffe.set_mode_gpu();

addpath('external/code3.2.1');
addpath(genpath('external/toolbox'));
ld = load(fullfile(model_dir, 'DeepCaltech_otfDetector'));
detector = ld.detector;
clear ld;

datasets = {'inria', 'eth', 'voc07', 'caltech', 'voc07'};

for data_num=1:length(datasets)
    data = jsondecode(fileread(['/home/yuliang/dataset/',char(datasets(data_num)),'.json']));

    data_type = string({'test'});
    im_names = [];  

    for k = 1:length(data_type)

        if data_type(k) == 'train' && length(data.train.images) ~= 0
            im_names = data.train.images;
            data.train.dets = cell(length(im_names),1);
        elseif data_type(k) == 'valid' && length(data.valid.images) ~= 0
            im_names = data.valid.images;
            data.valid.dets = cell(length(im_names),1);
        elseif data_type(k) == 'test' && length(data.test.images) ~= 0
            im_names = data.test.images;
            data.test.dets = cell(length(im_names),1);
        end

        for j = 1:length(im_names)
            im = imread(fullfile('/home/yuliang/dataset',im_names{j}));
            im = gpuArray(im);

            [boxes, scores] = proposal_im_detect_caltech(rpn_bf_model.conf_rpn, rpn_net, im);
            aboxes = [boxes, scores];
            [scores, indices] = sort(scores, 'descend');
            aboxes = aboxes(indices, :);
            aboxes = aboxes(1:1000, :);

            abox_size = size(aboxes);
            fprintf('RPN: width %d, height %d \n', abox_size(1), abox_size(2));

            if abox_size(1) == 0
                bbs = aboxes;
            else
                featmap_blobs_names = {'conv3_3', 'conv4_3_atrous'};
                featmap_blobs = cell(size(featmap_blobs_names));
                for i = 1:length(featmap_blobs_names);
                    featmap_blobs{i} = rpn_net.blobs(featmap_blobs_names{i});
                end
                feat1 = rois_get_features_from_featmap_ratio(rpn_bf_model.conf_bf, fast_rcnn_net, im, featmap_blobs, aboxes(:, 1:4), 2000, 1.0);
                % feat2 = rois_get_features_ratio(rpn_bf_model.conf_bf, caffe_net, im, aboxes(:, 1:4), 3000, 1.0);
                scores1 = adaBoostApply(feat1, detector.clf);
                % scores2 = adaBoostApply(feat2, detector.clf);

                % for i =1:length(scores)
                %     fprintf('score1: %f, score2: %f \n',scores1(i), scores2(i))
                % end

                bbs = [aboxes(:, 1:4) scores1];
                % [scores, indices] = sort(scores1, 'descend');
                [scores, indices] = sort(scores1);
                bbs = bbs(indices, :);
                bbs = bbs(1:200, :);
                bbs = rpn_boxes
            end
            bbs_size = size(bbs);

            fprintf('Final: width %d, height %d \n',bbs_size(1),bbs_size(2));

            fprintf('%s: %d / %d ===> %s (%dx%d)\n', data_type(k), j, length(im_names) , im_names{j}, size(im, 2), size(im, 1));
            
            if data_type(k) == 'train'
                data.train.dets{j,1} = aboxes;
            elseif data_type(k) == 'test'
                data.test.dets{j,1} = aboxes;
            elseif data_type(k) == 'valid'
                data.valid.dets{j,1} = aboxes;
            end
        end
    end

    save(['/home/yuliang/dataset/',char(datasets(data_num)),'_dets','.mat'], 'data');
end
caffe.reset_all(); 
clear mex;
end

function aboxes = boxes_filter(aboxes, per_nms_topN, nms_overlap_thres, after_nms_topN, use_gpu)
    % to speed up nms
    if per_nms_topN > 0
        aboxes = aboxes(1:min(length(aboxes), per_nms_topN), :);
    end
    % do nms
    if nms_overlap_thres > 0 && nms_overlap_thres < 1
        aboxes = aboxes(nms(aboxes, nms_overlap_thres, use_gpu), :);       
    end
    if after_nms_topN > 0
        aboxes = aboxes(1:min(length(aboxes), after_nms_topN), :);
    end
end

hello @zhangliliang ,I want to train the model on other dataset like INRIA or ETH , but there are only code for caltech , how should I add the training code? Thank you very much!

@zhangliliang Thank you for taking the time to read my questions!
The error is shown in the following image.
could you tell me where I need to modify the code?

Hi, I configured the project as your readme. Fortunately, I was able to run the testing demo successfully, but when I wanted to train the network, I had the following problems. I have never encountered a similar problem in matlab, so would I like to ask any suggestions from you?

There is the error:

RPN_BF startup done
GPU 1: free memory 11872763904
GPU 2: free memory 11872763904
Use GPU 1
Loading region proposals...done
Warrning: no windows proposal is loaded !
Reference to non-existent field 'sizes'.

Error in roidb_from_caltech (line 52)
  height = imdb.sizes(1,1);

Error in Dataset.caltech_trainval>@(x)x.roidb_func(x,false) (line 6)
        dataset.roidb_train   = cellfun(@(x) x.roidb_func(x, false), dataset.imdb_train, 'UniformOutput',
        false);

Error in Dataset.caltech_trainval (line 6)
        dataset.roidb_train   = cellfun(@(x) x.roidb_func(x, false), dataset.imdb_train, 'UniformOutput',
        false);

Error in script_rpn_pedestrian_VGG16_caltech (line 30)
dataset                     = Dataset.caltech_trainval(dataset, 'train');

I0111 12:25:53.896850 4324 net.cpp:235] This network produces output proposal_cls_prob
I0111 12:25:53.896872 4324 net.cpp:492] Collecting Learning Rate and Weight Decay.
I0111 12:25:53.896881 4324 net.cpp:247] Network initialization done.
I0111 12:25:53.896886 4324 net.cpp:248] Memory required for data: 162768800
I0111 12:25:53.986716 4324 net.cpp:746] Copying source layer conv1_1
I0111 12:25:53.986763 4324 net.cpp:746] Copying source layer relu1_1
I0111 12:25:53.986780 4324 net.cpp:746] Copying source layer conv1_2
I0111 12:25:53.986812 4324 net.cpp:746] Copying source layer relu1_2
I0111 12:25:53.986820 4324 net.cpp:746] Copying source layer pool1
I0111 12:25:53.986826 4324 net.cpp:746] Copying source layer conv2_1
I0111 12:25:53.986872 4324 net.cpp:746] Copying source layer relu2_1
I0111 12:25:53.986881 4324 net.cpp:746] Copying source layer conv2_2
I0111 12:25:53.986966 4324 net.cpp:746] Copying source layer relu2_2
I0111 12:25:53.986974 4324 net.cpp:746] Copying source layer pool2
I0111 12:25:53.986979 4324 net.cpp:746] Copying source layer conv3_1
I0111 12:25:53.987136 4324 net.cpp:746] Copying source layer relu3_1
I0111 12:25:53.987143 4324 net.cpp:746] Copying source layer conv3_2
I0111 12:25:53.987453 4324 net.cpp:746] Copying source layer relu3_2
I0111 12:25:53.987463 4324 net.cpp:746] Copying source layer conv3_3
I0111 12:25:53.987776 4324 net.cpp:746] Copying source layer relu3_3
I0111 12:25:53.987784 4324 net.cpp:746] Copying source layer pool3
I0111 12:25:53.987789 4324 net.cpp:746] Copying source layer conv4_1
I0111 12:25:53.988402 4324 net.cpp:746] Copying source layer relu4_1
I0111 12:25:53.988414 4324 net.cpp:746] Copying source layer conv4_2
I0111 12:25:53.989629 4324 net.cpp:746] Copying source layer relu4_2
I0111 12:25:53.989645 4324 net.cpp:746] Copying source layer conv4_3
I0111 12:25:53.990911 4324 net.cpp:746] Copying source layer relu4_3
I0111 12:25:53.990929 4324 net.cpp:746] Copying source layer pool4
I0111 12:25:53.990934 4324 net.cpp:746] Copying source layer conv5_1
I0111 12:25:53.992154 4324 net.cpp:746] Copying source layer relu5_1
I0111 12:25:53.992173 4324 net.cpp:746] Copying source layer conv5_2
I0111 12:25:53.993454 4324 net.cpp:746] Copying source layer relu5_2
I0111 12:25:53.993474 4324 net.cpp:746] Copying source layer conv5_3
I0111 12:25:53.994707 4324 net.cpp:746] Copying source layer relu5_3
I0111 12:25:53.994727 4324 net.cpp:746] Copying source layer conv_proposal1
I0111 12:25:53.995934 4324 net.cpp:746] Copying source layer relu_proposal1
I0111 12:25:53.995954 4324 net.cpp:746] Copying source layer conv_proposal1_relu_proposal1_0_split
I0111 12:25:53.995959 4324 net.cpp:746] Copying source layer proposal_cls_score
I0111 12:25:53.995973 4324 net.cpp:746] Copying source layer proposal_bbox_pred
I0111 12:25:53.995995 4324 net.cpp:746] Copying source layer proposal_cls_score_reshape
I0111 12:25:53.996001 4324 net.cpp:743] Ignoring source layer proposal_cls_score_reshape_proposal_cls_score_reshape_0_split
I0111 12:25:53.996007 4324 net.cpp:743] Ignoring source layer labels_reshape
I0111 12:25:53.996012 4324 net.cpp:743] Ignoring source layer labels_reshape_labels_reshape_0_split
I0111 12:25:53.996017 4324 net.cpp:743] Ignoring source layer labels_weights_reshape
I0111 12:25:53.996022 4324 net.cpp:743] Ignoring source layer loss
I0111 12:25:53.996027 4324 net.cpp:743] Ignoring source layer accuarcy
I0111 12:25:53.996031 4324 net.cpp:743] Ignoring source layer loss_bbox
I0111 12:25:53.997133 4324 net.cpp:42] Initializing net from parameters:
name: "VGG_ILSVRC_16"
input: "data1"
input: "data2"
input: "rois"
input_dim: 1
input_dim: 512
input_dim: 50
input_dim: 50
input_dim: 1
input_dim: 512
input_dim: 50
input_dim: 50
input_dim: 1
input_dim: 5
input_dim: 1
input_dim: 1
state {
phase: TEST
}
layer {
name: "roi_pool3"
type: "ROIPooling"
bottom: "data1"
bottom: "rois"
top: "roi_pool3"
roi_pooling_param {
pooled_h: 7
pooled_w: 7
spatial_scale: 0.25
}
}
layer {
name: "roi_pool4_3"
type: "ROIPooling"
bottom: "data2"
bottom: "rois"
top: "roi_pool4_3"
roi_pooling_param {
pooled_h: 7
pooled_w: 7
spatial_scale: 0.25
}
}
layer {
name: "roi_pool3_flat"
type: "Flatten"
bottom: "roi_pool3"
top: "roi_pool3_flat"
}
layer {
name: "roi_pool4_3_flat"
type: "Flatten"
bottom: "roi_pool4_3"
top: "roi_pool4_3_flat"
}
layer {
name: "concat_feat"
type: "Concat"
bottom: "roi_pool3_flat"
bottom: "roi_pool4_3_flat"
top: "concat_feat"
}
I0111 12:25:53.997277 4324 net.cpp:380] Input 0 -> data1
I0111 12:25:53.997287 4324 net.cpp:380] Input 1 -> data2
I0111 12:25:53.997294 4324 net.cpp:380] Input 2 -> rois
I0111 12:25:53.997303 4324 layer_factory.hpp:74] Creating layer rois_input_2_split
I0111 12:25:53.997311 4324 net.cpp:90] Creating Layer rois_input_2_split
I0111 12:25:53.997316 4324 net.cpp:420] rois_input_2_split <- rois
I0111 12:25:53.997323 4324 net.cpp:378] rois_input_2_split -> rois_input_2_split_0
I0111 12:25:53.997334 4324 net.cpp:378] rois_input_2_split -> rois_input_2_split_1
I0111 12:25:53.997344 4324 net.cpp:120] Setting up rois_input_2_split
I0111 12:25:53.997352 4324 net.cpp:127] Top shape: 1 5 1 1 (5)
I0111 12:25:53.997359 4324 net.cpp:127] Top shape: 1 5 1 1 (5)
I0111 12:25:53.997364 4324 layer_factory.hpp:74] Creating layer roi_pool3
I0111 12:25:53.997371 4324 net.cpp:90] Creating Layer roi_pool3
I0111 12:25:53.997377 4324 net.cpp:420] roi_pool3 <- data1
I0111 12:25:53.997383 4324 net.cpp:420] roi_pool3 <- rois_input_2_split_0
I0111 12:25:53.997390 4324 net.cpp:378] roi_pool3 -> roi_pool3
I0111 12:25:53.997397 4324 net.cpp:120] Setting up roi_pool3
I0111 12:25:53.997406 4324 roi_pooling_layer.cpp:44] Spatial scale: 0.25
I0111 12:25:53.997417 4324 net.cpp:127] Top shape: 1 512 7 7 (25088)
I0111 12:25:53.997423 4324 layer_factory.hpp:74] Creating layer roi_pool4_3
I0111 12:25:53.997429 4324 net.cpp:90] Creating Layer roi_pool4_3
I0111 12:25:53.997434 4324 net.cpp:420] roi_pool4_3 <- data2
I0111 12:25:53.997440 4324 net.cpp:420] roi_pool4_3 <- rois_input_2_split_1
I0111 12:25:53.997447 4324 net.cpp:378] roi_pool4_3 -> roi_pool4_3
I0111 12:25:53.997453 4324 net.cpp:120] Setting up roi_pool4_3
I0111 12:25:53.997459 4324 roi_pooling_layer.cpp:44] Spatial scale: 0.25
I0111 12:25:53.997467 4324 net.cpp:127] Top shape: 1 512 7 7 (25088)
I0111 12:25:53.997473 4324 layer_factory.hpp:74] Creating layer roi_pool3_flat
I0111 12:25:53.997480 4324 net.cpp:90] Creating Layer roi_pool3_flat
I0111 12:25:53.997486 4324 net.cpp:420] roi_pool3_flat <- roi_pool3
I0111 12:25:53.997493 4324 net.cpp:378] roi_pool3_flat -> roi_pool3_flat
I0111 12:25:53.997499 4324 net.cpp:120] Setting up roi_pool3_flat
I0111 12:25:53.997509 4324 net.cpp:127] Top shape: 1 25088 (25088)
I0111 12:25:53.997514 4324 layer_factory.hpp:74] Creating layer roi_pool4_3_flat
I0111 12:25:53.997521 4324 net.cpp:90] Creating Layer roi_pool4_3_flat
I0111 12:25:53.997527 4324 net.cpp:420] roi_pool4_3_flat <- roi_pool4_3
I0111 12:25:53.997534 4324 net.cpp:378] roi_pool4_3_flat -> roi_pool4_3_flat
I0111 12:25:53.997541 4324 net.cpp:120] Setting up roi_pool4_3_flat
I0111 12:25:53.997550 4324 net.cpp:127] Top shape: 1 25088 (25088)
I0111 12:25:53.997555 4324 layer_factory.hpp:74] Creating layer concat_feat
I0111 12:25:53.997562 4324 net.cpp:90] Creating Layer concat_feat
I0111 12:25:53.997568 4324 net.cpp:420] concat_feat <- roi_pool3_flat
I0111 12:25:53.997573 4324 net.cpp:420] concat_feat <- roi_pool4_3_flat
I0111 12:25:53.997581 4324 net.cpp:378] concat_feat -> concat_feat
I0111 12:25:53.997587 4324 net.cpp:120] Setting up concat_feat
I0111 12:25:53.997596 4324 net.cpp:127] Top shape: 1 50176 (50176)
I0111 12:25:53.997601 4324 net.cpp:194] concat_feat does not need backward computation.
I0111 12:25:53.997606 4324 net.cpp:194] roi_pool4_3_flat does not need backward computation.
I0111 12:25:53.997612 4324 net.cpp:194] roi_pool3_flat does not need backward computation.
I0111 12:25:53.997617 4324 net.cpp:194] roi_pool4_3 does not need backward computation.
I0111 12:25:53.997623 4324 net.cpp:194] roi_pool3 does not need backward computation.
I0111 12:25:53.997629 4324 net.cpp:194] rois_input_2_split does not need backward computation.
I0111 12:25:53.997635 4324 net.cpp:235] This network produces output concat_feat
I0111 12:25:53.997644 4324 net.cpp:492] Collecting Learning Rate and Weight Decay.
I0111 12:25:53.997650 4324 net.cpp:247] Network initialization done.
I0111 12:25:53.997654 4324 net.cpp:248] Memory required for data: 602152
F0111 12:25:55.269489 4324 relu_layer.cu:27] Check failed: error == cudaSuccess (18 vs. 0) invalid texture reference
*** Check failure stack trace: ***
Killed

When I tried to run 'script_rpn_bf_pedestrian_VGG16_caltech_demo.m' I got the above error. I had compiled MS version caffe and copy the mex file to the path. And I used Titan X as gpu and Matlab 2013b. Any idea for this error?? Thanks.

Hi,
Thank you for your sharing~
When I run script_rpn_bf_pedestrian_VGG16_caltech_demo to see the detection results,I have encountered the following problems.(Remarks: failed to use cudnn,so I modified the Makefile.config to disable the cudnn before compling).
I0108 18:23:22.370462 4073 net.cpp:90] Creating Layer roi_pool4_3_flat
I0108 18:23:22.370471 4073 net.cpp:420] roi_pool4_3_flat <- roi_pool4_3
I0108 18:23:22.370479 4073 net.cpp:378] roi_pool4_3_flat -> roi_pool4_3_flat
I0108 18:23:22.370491 4073 net.cpp:120] Setting up roi_pool4_3_flat
I0108 18:23:22.370501 4073 net.cpp:127] Top shape: 1 25088 (25088)
I0108 18:23:22.370507 4073 layer_factory.hpp:74] Creating layer concat_feat
I0108 18:23:22.370517 4073 net.cpp:90] Creating Layer concat_feat
I0108 18:23:22.370523 4073 net.cpp:420] concat_feat <- roi_pool3_flat
I0108 18:23:22.370532 4073 net.cpp:420] concat_feat <- roi_pool4_3_flat
I0108 18:23:22.370542 4073 net.cpp:378] concat_feat -> concat_feat
I0108 18:23:22.370550 4073 net.cpp:120] Setting up concat_feat
I0108 18:23:22.370561 4073 net.cpp:127] Top shape: 1 50176 (50176)
I0108 18:23:22.370568 4073 net.cpp:194] concat_feat does not need backward computation.
I0108 18:23:22.370576 4073 net.cpp:194] roi_pool4_3_flat does not need backward computation.
I0108 18:23:22.370584 4073 net.cpp:194] roi_pool3_flat does not need backward computation.
I0108 18:23:22.370590 4073 net.cpp:194] roi_pool4_3 does not need backward computation.
I0108 18:23:22.370597 4073 net.cpp:194] roi_pool3 does not need backward computation.
I0108 18:23:22.370605 4073 net.cpp:194] rois_input_2_split does not need backward computation.
I0108 18:23:22.370612 4073 net.cpp:235] This network produces output concat_feat
I0108 18:23:22.370625 4073 net.cpp:492] Collecting Learning Rate and Weight Decay.
I0108 18:23:22.370632 4073 net.cpp:247] Network initialization done.
I0108 18:23:22.370640 4073 net.cpp:248] Memory required for data: 602152
F0108 18:23:27.069090 4073 syncedmem.cpp:51] Check failed: error == cudaSuccess (2 vs. 0) out of memory
*** Check failure stack trace: ***
Killed

Is it caused by the disable cudnn?How should I fix this problem? Any suggestions are very grateful!Thank you very much.

I1114 13:08:42.275104 31069 sgd_solver.cpp:106] Iteration 150, lr = 5e-05
I1114 13:09:43.542522 31069 solver.cpp:236] Iteration 200, loss = 5.62384
I1114 13:09:43.542560 31069 solver.cpp:252] Train net output #0: accuracy_1_5x5 = 0.881346
I1114 13:09:43.542567 31069 solver.cpp:252] Train net output #1: accuracy_1_5x5 = 0.0714286
I1114 13:09:43.542572 31069 solver.cpp:252] Train net output #2: accuracy_1_7x7 = 0.817905
I1114 13:09:43.542574 31069 solver.cpp:252] Train net output #3: accuracy_1_7x7 = 0.454545
I1114 13:09:43.542578 31069 solver.cpp:252] Train net output #4: accuracy_2_5x5 = 0.780908
I1114 13:09:43.542582 31069 solver.cpp:252] Train net output #5: accuracy_2_5x5 = 1
I1114 13:09:43.542587 31069 solver.cpp:252] Train net output #6: accuracy_2_7x7 = 0.89241
I1114 13:09:43.542590 31069 solver.cpp:252] Train net output #7: accuracy_2_7x7 = -1
I1114 13:09:43.542593 31069 solver.cpp:252] Train net output #8: accuracy_3_5x5 = 0.841772
I1114 13:09:43.542598 31069 solver.cpp:252] Train net output #9: accuracy_3_5x5 = 0
I1114 13:09:43.542600 31069 solver.cpp:252] Train net output #10: accuracy_3_7x7 = 0.731013
I1114 13:09:43.542604 31069 solver.cpp:252] Train net output #11: accuracy_3_7x7 = 1
I1114 13:09:43.542608 31069 solver.cpp:252] Train net output #12: accuracy_4_5x5 = 0.955696
I1114 13:09:43.542611 31069 solver.cpp:252] Train net output #13: accuracy_4_5x5 = 0.5
I1114 13:09:43.542628 31069 solver.cpp:252] Train net output #14: boxiou_1_5x5 = 0.596698
I1114 13:09:43.542631 31069 solver.cpp:252] Train net output #15: boxiou_1_7x7 = 0.54248
I1114 13:09:43.542635 31069 solver.cpp:252] Train net output #16: boxiou_2_5x5 = 0.59975
I1114 13:09:43.542639 31069 solver.cpp:252] Train net output #17: boxiou_2_7x7 = -1
I1114 13:09:43.542642 31069 solver.cpp:252] Train net output #18: boxiou_3_5x5 = 0.547801
I1114 13:09:43.542646 31069 solver.cpp:252] Train net output #19: boxiou_3_7x7 = 0.58118
I1114 13:09:43.542650 31069 solver.cpp:252] Train net output #20: boxiou_4_5x5 = 0.600571
I1114 13:09:43.542675 31069 solver.cpp:252] Train net output #21: loss_1_5x5 = 1.26751 (* 0.9 = 1.14076 loss)
I1114 13:09:43.542682 31069 solver.cpp:252] Train net output #22: loss_1_5x5 = 0.000743121 (* 0.9 = 0.000668809 loss)
I1114 13:09:43.542688 31069 solver.cpp:252] Train net output #23: loss_1_7x7 = 3.04133 (* 0.9 = 2.7372 loss)
I1114 13:09:43.542695 31069 solver.cpp:252] Train net output #24: loss_1_7x7 = 0.000900266 (* 0.9 = 0.00081024 loss)
I1114 13:09:43.542701 31069 solver.cpp:252] Train net output #25: loss_2_5x5 = 0.374449 (* 1 = 0.374449 loss)
I1114 13:09:43.542706 31069 solver.cpp:252] Train net output #26: loss_2_5x5 = 0.000737343 (* 1 = 0.000737343 loss)
I1114 13:09:43.542711 31069 solver.cpp:252] Train net output #27: loss_2_7x7 = 0.183456 (* 1 = 0.183456 loss)
I1114 13:09:43.542717 31069 solver.cpp:252] Train net output #28: loss_2_7x7 = 0 (* 1 = 0 loss)
I1114 13:09:43.542722 31069 solver.cpp:252] Train net output #29: loss_3_5x5 = 0.471607 (* 1 = 0.471607 loss)
I1114 13:09:43.542728 31069 solver.cpp:252] Train net output #30: loss_3_5x5 = 0.00156796 (* 1 = 0.00156796 loss)
I1114 13:09:43.542733 31069 solver.cpp:252] Train net output #31: loss_3_7x7 = 0.351334 (* 1 = 0.351334 loss)
I1114 13:09:43.542739 31069 solver.cpp:252] Train net output #32: loss_3_7x7 = 0.000982699 (* 1 = 0.000982699 loss)
I1114 13:09:43.542745 31069 solver.cpp:252] Train net output #33: loss_4_5x5 = 0.359949 (* 1 = 0.359949 loss)
I1114 13:09:43.542750 31069 solver.cpp:252] Train net output #34: loss_4_5x5 = 0.000321437 (* 1 = 0.000321437 loss)

when I run script_rpn_pedestrian_VGG16_caltech.m, I get the error:
I0905 13:10:57.821876 2060 net.cpp:194] relu_proposal1 does not need backward computation.
I0905 13:10:57.821879 2060 net.cpp:194] conv_proposal1 does not need backward computation.
I0905 13:10:57.821884 2060 net.cpp:194] relu5 does not need backward computation.
I0905 13:10:57.821888 2060 net.cpp:194] conv5 does not need backward computation.
I0905 13:10:57.821893 2060 net.cpp:194] relu4 does not need backward computation.
I0905 13:10:57.821899 2060 net.cpp:194] conv4 does not need backward computation.
I0905 13:10:57.821905 2060 net.cpp:194] relu3 does not need backward computation.
I0905 13:10:57.821910 2060 net.cpp:194] conv3 does not need backward computation.
I0905 13:10:57.821916 2060 net.cpp:194] pool2 does not need backward computation.
I0905 13:10:57.821923 2060 net.cpp:194] norm2 does not need backward computation.
I0905 13:10:57.821928 2060 net.cpp:194] relu2 does not need backward computation.
I0905 13:10:57.821933 2060 net.cpp:194] conv2 does not need backward computation.
I0905 13:10:57.821939 2060 net.cpp:194] pool1 does not need backward computation.
I0905 13:10:57.821944 2060 net.cpp:194] norm1 does not need backward computation.
I0905 13:10:57.821950 2060 net.cpp:194] relu1 does not need backward computation.
I0905 13:10:57.821955 2060 net.cpp:194] conv1 does not need backward computation.
I0905 13:10:57.821960 2060 net.cpp:235] This network produces output proposal_bbox_pred
I0905 13:10:57.821965 2060 net.cpp:235] This network produces output proposal_cls_prob
I0905 13:10:57.821980 2060 net.cpp:492] Collecting Learning Rate and Weight Decay.
I0905 13:10:57.821990 2060 net.cpp:247] Network initialization done.
I0905 13:10:57.821995 2060 net.cpp:248] Memory required for data: 21358056
F0905 13:10:57.868927 2060 relu_layer.cu:27] Check failed: error == cudaSuccess (18 vs. 0) invalid texture reference
*** Check failure stack trace: ***
Killed

Hi,

Thanks for the effort. I tried to make this caffe on my machine, but it keeps showing error:

CXX/LD -o .build_release/examples/cifar10/convert_cifar_data.bin
.build_release/lib/libcaffe.so: undefined reference to cv::_OutputArray::_OutputArray(cv::Mat&)' .build_release/lib/libcaffe.so: undefined reference to cv::imencode(std::string const&, cv::_InputArray const&, std::vector<unsigned char, std::allocator >&, std::vector<int, std::allocator > const&)'
.build_release/lib/libcaffe.so: undefined reference to cv::_InputArray::_InputArray(cv::Mat const&)' .build_release/lib/libcaffe.so: undefined reference to cv::imread(std::string const&, int)'
.build_release/lib/libcaffe.so: undefined reference to vtable for cv::_InputArray' collect2: error: ld returned 1 exit status make: *** [.build_release/tools/convert_imageset.bin] Error 1 make: *** Waiting for unfinished jobs.... .build_release/lib/libcaffe.so: undefined reference to cv::_OutputArray::_OutputArray(cv::Mat&)'
.build_release/lib/libcaffe.so: undefined reference to cv::imencode(std::string const&, cv::_InputArray const&, std::vector<unsigned char, std::allocator<unsigned char> >&, std::vector<int, std::allocator<int> > const&)' .build_release/lib/libcaffe.so: undefined reference to cv::_InputArray::_InputArray(cv::Mat const&)'
.build_release/lib/libcaffe.so: undefined reference to cv::imread(std::string const&, int)' .build_release/lib/libcaffe.so: undefined reference to vtable for cv::_InputArray'
collect2: error: ld returned 1 exit status
make: *** [.build_release/tools/compute_image_mean.bin] Error 1
.build_release/lib/libcaffe.so: undefined reference to cv::_OutputArray::_OutputArray(cv::Mat&)' .build_release/lib/libcaffe.so: undefined reference to cv::imencode(std::string const&, cv::_InputArray const&, std::vector<unsigned char, std::allocator >&, std::vector<int, std::allocator > const&)'
.build_release/lib/libcaffe.so: undefined reference to cv::_InputArray::_InputArray(cv::Mat const&)' .build_release/lib/libcaffe.so: undefined reference to cv::imread(std::string const&, int)'
.build_release/lib/libcaffe.so: undefined reference to vtable for cv::_InputArray' collect2: error: ld returned 1 exit status make: *** [.build_release/tools/caffe.bin] Error 1 .build_release/lib/libcaffe.so: undefined reference to cv::_OutputArray::_OutputArray(cv::Mat&)'
.build_release/lib/libcaffe.so: undefined reference to cv::imencode(std::string const&, cv::_InputArray const&, std::vector<unsigned char, std::allocator<unsigned char> >&, std::vector<int, std::allocator<int> > const&)' .build_release/lib/libcaffe.so: undefined reference to cv::_InputArray::_InputArray(cv::Mat const&)'
.build_release/lib/libcaffe.so: undefined reference to cv::imread(std::string const&, int)' .build_release/lib/libcaffe.so: undefined reference to vtable for cv::_InputArray'
collect2: error: ld returned 1 exit status
make: *** [.build_release/tools/extract_features.bin] Error 1
.build_release/lib/libcaffe.so: undefined reference to cv::_OutputArray::_OutputArray(cv::Mat&)' .build_release/lib/libcaffe.so: undefined reference to cv::imencode(std::string const&, cv::_InputArray const&, std::vector<unsigned char, std::allocator >&, std::vector<int, std::allocator > const&)'
.build_release/lib/libcaffe.so: undefined reference to cv::_InputArray::_InputArray(cv::Mat const&)' .build_release/lib/libcaffe.so: undefined reference to cv::imread(std::string const&, int)'
.build_release/lib/libcaffe.so: undefined reference to vtable for cv::_InputArray' collect2: error: ld returned 1 exit status make: *** [.build_release/tools/upgrade_net_proto_binary.bin] Error 1 .build_release/lib/libcaffe.so: undefined reference to cv::_OutputArray::_OutputArray(cv::Mat&)'
.build_release/lib/libcaffe.so: undefined reference to cv::imencode(std::string const&, cv::_InputArray const&, std::vector<unsigned char, std::allocator<unsigned char> >&, std::vector<int, std::allocator<int> > const&)' .build_release/lib/libcaffe.so: undefined reference to cv::_InputArray::_InputArray(cv::Mat const&)'
.build_release/lib/libcaffe.so: undefined reference to cv::imread(std::string const&, int)' .build_release/lib/libcaffe.so: undefined reference to vtable for cv::_InputArray'
collect2: error: ld returned 1 exit status
make: *** [.build_release/examples/cifar10/convert_cifar_data.bin] Error 1
.build_release/lib/libcaffe.so: undefined reference to cv::_OutputArray::_OutputArray(cv::Mat&)' .build_release/lib/libcaffe.so: undefined reference to cv::imencode(std::string const&, cv::_InputArray const&, std::vector<unsigned char, std::allocator >&, std::vector<int, std::allocator > const&)'
.build_release/lib/libcaffe.so: undefined reference to cv::_InputArray::_InputArray(cv::Mat const&)' .build_release/lib/libcaffe.so: undefined reference to cv::imread(std::string const&, int)'
.build_release/lib/libcaffe.so: undefined reference to vtable for cv::_InputArray' collect2: error: ld returned 1 exit status make: *** [.build_release/tools/upgrade_net_proto_text.bin] Error 1 .build_release/lib/libcaffe.so: undefined reference to cv::_OutputArray::_OutputArray(cv::Mat&)'
.build_release/lib/libcaffe.so: undefined reference to cv::imencode(std::string const&, cv::_InputArray const&, std::vector<unsigned char, std::allocator<unsigned char> >&, std::vector<int, std::allocator<int> > const&)' .build_release/lib/libcaffe.so: undefined reference to cv::_InputArray::_InputArray(cv::Mat const&)'
.build_release/lib/libcaffe.so: undefined reference to cv::imread(std::string const&, int)' .build_release/lib/libcaffe.so: undefined reference to vtable for cv::_InputArray'
collect2: error: ld returned 1 exit status
make: *** [.build_release/examples/siamese/convert_mnist_siamese_data.bin] Error 1

I learnt someone encountering the same problem because they were using OPENCV3.0 and had to set the OPENCV_VERSION variable or add opencv_imgcodecs to the LIBRARIES. However I'm quite sure I'm using OPENCV 2.4.11 now, and even I play these tricks the error still shows.

Can anyone provide some suggestion? Appreciate it.

the error reads like this:
Loading region proposals...done
Warrning: no windows proposal is loaded !
引用了不存在的字段 'sizes'。

出错 roidb_from_caltech (line 52)
height = imdb.sizes(1,1);

出错 Dataset.caltech_trainval>@(x)x.roidb_func(x,false) (line 6)
dataset.roidb_train = cellfun(@(x) x.roidb_func(x, false), dataset.imdb_train, 'UniformOutput', false);

出错 Dataset.caltech_trainval (line 6)
dataset.roidb_train = cellfun(@(x) x.roidb_func(x, false), dataset.imdb_train, 'UniformOutput', false);

出错 script_rpn_pedestrian_VGG16_caltech (line 30)
dataset = Dataset.caltech_trainval(dataset, 'train');
Here is the screenshot of the dataset, have I got something wrong?
Thanks a lot! @zhangliliang

Hello!

Thanks for sharing the caffe which added "a trous" trick support.

I can install your caffe on linux successfully, but something goes wrong on windows.

The errors as follow:

错误 11771 error C2039: “has_filter_stride_h”: 不是“caffe::ConvolutionParameter”的成员 \src\caffe\layers\im2col_layer.cpp 28 1 caffe
错误 11774 error C2039: “filter_stride”: 不是“caffe::ConvolutionParameter”的成员 \src\caffe\layers\im2col_layer.cpp 53 1 caffe

I can install caffe that support for Faster RCNN on windows successfully, but I use same way to install your caffe, those errors just come out.

I'll be grateful if you can help me.

thanks!

Hi @zhangliliang
Sorry to bother you again，but I'm wondering how you dealed with ground truths annotated as "people" or "person?", since these ground truths may introduce some noise during training. I prefer simply ignore them.
Thanks a lot!

My computer has the nvidia graphic card which is GTX980
but when I run the demo you given, always out of memory,can you give me some advice to solve the problem ?
Thank you

Hello, I'm sorry to disturb you. I'm trying to re-implement the code and I have some questions for compiling and training the RPN.

For the compiling part, I'm not sure if I'm doing it right. First, I downloaded and compiled the Caffe, which is mentiond in the Requirements section. Second, I run the rpn_bf_build.m and startup.m and also the script_rpn_bf_pedestrian_VGG16_caltech_demo.m. The demo seems succeed, but I'm wondering is that right or not for using rpn_bf_build.m instead of faster_rcnn_build.m.

For training the RPN part, I have successfully extract images and create the annos. I also downloaded the pre-trained model from website, and unzip in the repo folder. But I have a problem on running the script_rpn_pedestrian_VGG16_caltech.m, which is that my computer will crash without warning at "preparing data" part after running for 10 or 20 minutes. My environment is ubuntu 14.04 and Nvidia Titan X, with cuda 7.5 and cudnn v3. I have noticed that even the code shows "use GPU 1", which is Titan X, the loading of my GPU memory won't increase(I use the "nvidia-smi" command in terminal to see the memory usage), but the RAM loading(system monitor) of my pc will increase and then it crashed.

Could you please give some advice? Thanks a lot!

Subscripted assignment dimension mismatch.

Error in DeepTrain_otf_trans_ratio>sampleWins (line 454)
bg_feat(bg_feat_idx:bg_feat_idx+length(sel_idx)-1, :) = sel_feat;

Error in DeepTrain_otf_trans_ratio (line 148)
[X0, X0_score, sel_idxes] = sampleWins( detector, stage, 0 );

Error in script_rpn_bf_pedestrian_VGG16_caltech (line 146)
detector = DeepTrain_otf_trans_ratio( opts );

What do I need to change if I can implement in python

Im tried to build rpn-bf project(this repository) after training, the result is it can not classification any image . so I change to some parameters but i can not solve this.

So , i request to final rpn result in this github , anyone who have that contact me

Hello! Thank you for making this great code available!

Why is running the training code gives ~10 MR (I get 10.03), but the paper claims 9.6? Are there any differences in the paper vs. github implmentation? Are the RPN networks used the same, so that maybe just the random seed in BF is the cause, or is there something else? Thanks!

When running script_rpn_bf_pedestrian_VGG16_caltech, I got the following error:
`Error using caffe_
Protobuf : Error parsing text-format caffe.NetParameter: 1:1: Expected
identifier. . at google/protobuf/text_format.cc Line 245

Error in caffe.get_net (line 27)
hNet = caffe_('get_net', model_file, phase_name);

Error in caffe.Net (line 33)
self = caffe.get_net(varargin{:});

Error in script_rpn_bf_pedestrian_VGG16_caltech (line 81)
caffe_net = caffe.Net(BF_prototxt_path, 'test');`

As you mentioned in #2, the file is a text file and can be opened via gedit or vim. I tried that and found that it cannot be displayed normally. Can you check if the file is corrupted?

First of all, thank you for your job!
However, when I run script_rpn_pedestrian_VGG16_caltech.m , my computer jamed and can not do anything, staying in the stage "stage one RPN"! Then, I waited for about 1 hour, it's still jamed.
So I just want to ask the minimum requirement about machine.
Here is my machine : One Titan x GPU; 32G memory. Is enough for this experiment?
Thank you very much!

hi,sorry to bother you!
I am very confused about the dataset in the step below
Download the annotations and videos in Caltech Pedestrian Dataset and put them in the proper folder follow the instruction in the website.

what is a proper folder ?

could you give me some more detail,I cannot find the detail in the website!
Thank you !!

Hi,thank you for your job and sorry to disturb you. I have some confusion about the requirements of this test. I see your environment configuration is Ubuntu 14..cuda 7.5. Will Ubuntu 16.04, cuda 8.0, cudnn v5, such an environment to run your program can be compatible?

Hello,
sorry to disturb you ,have you make the test for uisng the faster-rcnn in Caltech。i know your paper'title is "Is faster rcnn done well for the .",do you make the test. If you have ,would you copy an code to the github,Thank you !

Hi Zhang,
I was trying to reproduce the result on Caltech test set, using the pretrained model given for the demo . I made a few modefications on script_rpn_pedestrian_VGG16_caltech.m, simply skip the training part so that I can use the pretrained model to do evaluation directly. I got only 10.4% MR on Caltech test set, compared to 9.6% in your paper. Is there any difference between the pretrained model given for the demo and the one you used in the paper? Or any other explanation for my result?
Thanks a lot!

Hi @zhangliliang,
I notice that there are a number of images containing no ground truth objects in the caltech training set. I'm just wondering how you dealed with them? Did you just discard them?
Thanks!

Hello!

When compiling the specified Caffe version, I got the following errors:
/tmp/mex_9704849759849136_22812/caffe_.o: In functionmx_mat_to_blob(mxArray_tag const_, caffe::Blob, WhichMemory)':
caffe.cpp:(.text+0x1674): undefined reference to mxIsGPUArray' caffe_.cpp:(.text+0x167d): undefined reference tomxInitGPU'
caffe_.cpp:(.text+0x1685): undefined reference to mxGPUCreateFromMxArray' caffe_.cpp:(.text+0x1690): undefined reference tomxGPUGetDataReadOnly'
caffe_.cpp:(.text+0x16a3): undefined reference to mxGPUGetNumberOfElements' caffe_.cpp:(.text+0x1854): undefined reference tomxIsGPUArray'
caffe_.cpp:(.text+0x1860): undefined reference to mxGPUDestroyGPUArray' /tmp/mex_9704849759849136_22812/caffe_.o: In functionblob_set_data(int, mxArray_tag**, int, mxArray_tag const**)':
caffe_.cpp:(.text+0x18f8): undefined reference to mxIsGPUArray' collect2: error: ld returned 1 exit status

Do you have any idea on how to fix them? Thanks.

求助。。 我换了自己的数据库训练的时候出现:
Error using arrayfun
Sparse Arrays are not supported.
Error in proposal_train_caltech>generate_random_minibatch(line 195)

弄了好久了，就是找不到问题在哪，我把数据库做成了和caltech一样的格式。
打扰您了。

When I am running this file : 'script_rpn_pedestrian_VGG16_caltech' I am getting matlab is crashed and getting this error in the terminal:
Check failed: fd != -1 (-1 vs. -1) File not found: .\models\rpn_prototxts\vgg_16layers_conv3_1\train_val.prototxt
*** Check failure stack trace: ***
But I have this file "train_val.prototxt" in above mentioned path
Any one encountered this kind of error? Please provide a path out of this.

I run the script_rpn_bf_pedestrian_VGG16_caltech_demo.m with some images, it works well, then I change the images name and run again, the program will fail， even if I do not change the names, the second times it still fails.
The error in matlab command window is
Caught "std::exception" Exception message is
CHECK failed , generated_database_->Add(encode_file_descripor, size);

I do not know what happened, Does anyone know why？

Hello @zhangliliang!

When I follow the Training on Caltech (RPN) instruction, I meet this problem:

Preparing the results for Caltech evaluation ...Done.Cannot find an exact (case-sensitive) match for 'dbInfo'

The closest match is: dbinfo in /usr/local/MATLAB/R2014/toolbox/wavelet/wavelet/dbinfo.m

Error in dbEval_RPNBF (line 102)
  [~,set]=dbInfo(dataName); set=['/set' int2str2(set(1),2)];

Error in Faster_RCNN_Train.do_proposal_test_caltech (line 62)
    dbEval_RPNBF;

Error in script_rpn_pedestrian_VGG16_caltech (line 53)
Faster_RCNN_Train.do_proposal_test_caltech(conf_proposal, model.stage1_rpn, dataset.imdb_test, dataset.roidb_test, cache_name, method_name);

Error in run (line 63)
evalin('caller', [script ';']);

Do you have any idea to fix it? Thanks a lot!

The error looks like this:

......
Pos: 42500 / 42782 (select 30434 gt)
Pos: 42600 / 42782 (select 30472 gt)
Pos: 42700 / 42782 (select 30472 gt)
Select pos num: 30472

done (time=2315s).

Training stage 0
Neg: 2476 / 30000 (from 100 images)
Neg: 4976 / 30000 (from 200 images)
Neg: 7476 / 30000 (from 300 images)
Neg: 9976 / 30000 (from 400 images)
Neg: 12476 / 30000 (from 500 images)
Neg: 14976 / 30000 (from 600 images)
Neg: 17476 / 30000 (from 700 images)
Neg: 19976 / 30000 (from 800 images)
Neg: 22476 / 30000 (from 900 images)
Neg: 24976 / 30000 (from 1000 images)
Neg: 27476 / 30000 (from 1100 images)
Neg: 29976 / 30000 (from 1200 images)
Select neg num: 30000 from 1200 images
done (time=309s).
Training AdaBoost: nWeak= 64 nFtrs=37632 pos=30472 neg=30000
Invalid MEX-file
'/media/sjtu/0CE436CFE436BB32/code/RPN_BF-RPN-pedestrian/external/toolbox/classify/private/binaryTreeTrain1.mexa64': dlopen:
cannot load any more object with static TLS

出错 binaryTreeTrain (line 106)
[errsSt,thrsSt] = binaryTreeTrain1(X0,X1,single(wts0/w),...

出错 adaBoostTrain_trans (line 104)
[tree,data,err]=binaryTreeTrain(data,pTree);

出错 DeepTrain_otf_trans_ratio (line 185)
detector.clf = adaBoostTrain_trans(X0,X0_score,X1,X1_score,detector.opts.pBoost);

出错 script_rpn_bf_pedestrian_VGG16_caltech (line 146)
detector = DeepTrain_otf_trans_ratio( opts );