Faster RCNN训练（Matlab版本）结果

Faster RCNN训练（Matlab版本）结果：

matfaster_rcnn: test (voc_2007_trainval) 187/200  time: 0.029s
matfaster_rcnn: test (voc_2007_trainval) 188/200  time: 0.027s
......
matfaster_rcnn: test (voc_2007_trainval) 200/200  time: 0.028s
Cleared 0 solvers and 1 stand-alone nets
aver_boxes_num = 8201, select top 2000
opts:
      cache_name: 'faster_rcnn_VOC2007_ZF_top-1_nms0_7_top2000_stage2_rpn'
            conf: [1x1 struct]
            imdb: [1x1 struct]
    net_def_file: '/home/faster_rcnn/models/rpn_prototxts/ZF_fc6/test.prototxt'
        net_file: '/home/faster_rcnn/output/rpn_cachedir/faster_rcnn_VOC2007_ZF_top-1_nms0_7_top2000_stage2_rpn/voc_2007_trainval/final'
          suffix: ''

conf:
                      batch_size: 256
                    bg_thresh_hi: 0.3000
                    bg_thresh_lo: 0
                       bg_weight: 1
         drop_boxes_runoff_image: 1
                     feat_stride: 16
                     fg_fraction: 0.5000
                       fg_thresh: 0.7000
                     image_means: [224x224x3 single]
                   ims_per_batch: 1
                        max_size: 1000
                        rng_seed: 6
                          scales: 600
                  target_only_gt: 1
                     test_binary: 0
    test_drop_boxes_runoff_image: 0
                   test_max_size: 1000
               test_min_box_size: 16
                        test_nms: 0.3000
                     test_scales: 600
                     use_flipped: 1
                         use_gpu: 1
                         anchors: [9x4 double]
                output_width_map: [901x1 containers.Map]
               output_height_map: [901x1 containers.Map]

matfaster_rcnn: test (voc_2007_test) 1/100  time: 0.033s
matfaster_rcnn: test (voc_2007_test) 2/100  time: 0.027s
......

matfaster_rcnn: test (voc_2007_test) 100/100  time: 0.027s
Cleared 0 solvers and 1 stand-alone nets
aver_boxes_num = 8206, select top 2000

***************
stage two fast rcnn
***************
conf:
       batch_size: 128
      bbox_thresh: 0.5000
     bg_thresh_hi: 0.5000
     bg_thresh_lo: 0.1000
      fg_fraction: 0.2500
        fg_thresh: 0.5000
      image_means: [224x224x3 single]
    ims_per_batch: 2
         max_size: 1000
         rng_seed: 6
           scales: 600
      test_binary: 0
    test_max_size: 1000
         test_nms: 0.3000
      test_scales: 600
      use_flipped: 1
          use_gpu: 1

opts:
           cache_name: 'faster_rcnn_VOC2007_ZF_top-1_nms0_7_top2000_stage2_fast_rcnn'
                 conf: [1x1 struct]
               do_val: 1
           imdb_train: {[1x1 struct]}
             imdb_val: [1x1 struct]
             net_file: '/home/faster_rcnn/output/fast_rcnn_cachedir/faster_rcnn_VOC2007_ZF_top-1_nms0_7_top2000_stage1_fast_rcnn/voc_2007_train...'
          roidb_train: {[1x1 struct]}
            roidb_val: [1x1 struct]
    snapshot_interval: 10000
      solver_def_file: '/home/faster_rcnn/models/fast_rcnn_prototxts/ZF_fc6/solver_30k40k.prototxt'
         val_interval: 2500
            val_iters: 30

Preparing training data...Done.
Preparing validation data...Done.

------------------------- Iteration 2500 -------------------------
Training : error 0.14, loss (cls 0.331, reg 0.301)
Testing  : error 0.133, loss (cls 0.313, reg 0.281)

------------------------- Iteration 5000 -------------------------
Training : error 0.136, loss (cls 0.315, reg 0.291)
Testing  : error 0.122, loss (cls 0.303, reg 0.296)

------------------------- Iteration 7500 -------------------------
Training : error 0.131, loss (cls 0.305, reg 0.289)
Testing  : error 0.117, loss (cls 0.281, reg 0.257)

------------------------- Iteration 10000 -------------------------
Training : error 0.128, loss (cls 0.297, reg 0.284)
Testing  : error 0.125, loss (cls 0.29, reg 0.27)
Saved as /home/faster_rcnn/output/fast_rcnn_cachedir/faster_rcnn_VOC2007_ZF_top-1_nms0_7_top2000_stage2_fast_rcnn/voc_2007_trainval/iter_10000

......

------------------------- Iteration 37500 -------------------------
Training : error 0.0953, loss (cls 0.215, reg 0.255)
Testing  : error 0.0917, loss (cls 0.198, reg 0.246)
Saved as /home/faster_rcnn/output/fast_rcnn_cachedir/faster_rcnn_VOC2007_ZF_top-1_nms0_7_top2000_stage2_fast_rcnn/voc_2007_trainval/iter_40000
Saved as /home/faster_rcnn/output/fast_rcnn_cachedir/faster_rcnn_VOC2007_ZF_top-1_nms0_7_top2000_stage2_fast_rcnn/voc_2007_trainval/final
Cleared 1 solvers and 0 stand-alone nets

***************
final test
***************
aver_boxes_num = 2780, select top 300
opts:
      cache_name: 'faster_rcnn_VOC2007_ZF_top-1_nms0_7_top2000_stage2_fast_rcnn'
            conf: [1x1 struct]
    ignore_cache: 0
            imdb: [1x1 struct]
    net_def_file: '/home/faster_rcnn/models/fast_rcnn_prototxts/ZF_fc6/test.prototxt'
        net_file: '/home/faster_rcnn/output/fast_rcnn_cachedir/faster_rcnn_VOC2007_ZF_top-1_nms0_7_top2000_stage2_fast_rcnn/voc_2007_trainval/f...'
           roidb: [1x1 struct]
          suffix: ''

conf:
       batch_size: 128
      bbox_thresh: 0.5000
     bg_thresh_hi: 0.5000
     bg_thresh_lo: 0.1000
      fg_fraction: 0.2500
        fg_thresh: 0.5000
      image_means: [224x224x3 single]
    ims_per_batch: 2
         max_size: 1000
         rng_seed: 6
           scales: 600
      test_binary: 0
    test_max_size: 1000
         test_nms: 0.3000
      test_scales: 600
      use_flipped: 1
          use_gpu: 1

matfaster_rcnn: test (voc_2007_test) 1/100  time: 0.034s
......

matfaster_rcnn: test (voc_2007_test) 100/100  time: 0.033s
    0.5057

test all images in 3.363368 seconds.
Cleared 0 solvers and 1 stand-alone nets
IdleTimeout has been reached.
Parallel pool using the 'local' profile is shutting down.

do_eval =

     1

obj: pr: load: 29/100
obj: pr: load: 54/100
obj: pr: load: 77/100
obj: pr: load: 93/100
!!! obj : 0.3506 0.3117

~~~~~~~~~~~~~~~~~~~~
Results:
   35.0550

   35.0550

~~~~~~~~~~~~~~~~~~~~
Cleared 0 solvers and 2 stand-alone nets
please modify detection_test.prototxt file for sharing conv layers with proposal model (delete layers until relu5)
>> 

转载请注明：http://blog.csdn.net/forest_world