|
12 | 12 | switch data_name
|
13 | 13 | case 'ADE20K'
|
14 | 14 | isVal = true; %evaluation on valset
|
15 |
| - step = 2000; %equals to number of images divide num of GPUs in testing e.g. 500=2000/4 |
| 15 | + step = 500; %equals to number of images divide num of GPUs in testing e.g. 500=2000/4 |
16 | 16 | data_root = '/data2/hszhao/dataset/ADEChallengeData2016'; %root path of dataset
|
17 | 17 | eval_list = 'list/ADE20K_val.txt'; %evaluation list, refer to lists in folder 'samplelist'
|
18 |
| - eval_list = 'validation.txt'; %evaluation list, refer to lists in folder 'samplelist' |
19 |
| - save_root = 'mc_result/ADE20K/val/pspnet50_473_g0/'; %root path to store the result image |
| 18 | + save_root = 'mc_result/ADE20K/val/pspnet50_473/'; %root path to store the result image |
20 | 19 | model_weights = 'model/pspnet50_ADE20K.caffemodel';
|
21 | 20 | model_deploy = 'prototxt/pspnet50_ADE20K_473.prototxt';
|
22 | 21 | fea_cha = 150; %number of classes
|
|
26 | 25 | data_colormap = 'color150.mat'; %color map
|
27 | 26 | case 'VOC2012'
|
28 | 27 | isVal = false; %evaluation on testset
|
29 |
| - step = 364; |
| 28 | + step = 364; %364=1456/4 |
30 | 29 | data_root = '/data2/hszhao/dataset/VOC2012';
|
31 | 30 | eval_list = 'list/VOC2012_test.txt';
|
32 |
| - eval_list = 'list/test.txt'; |
33 | 31 | save_root = 'mc_result/VOC2012/test/pspnet101_473/';
|
34 | 32 | model_weights = 'model/pspnet101_VOC2012.caffemodel';
|
35 | 33 | model_deploy = 'prototxt/pspnet101_VOC2012_473.prototxt';
|
|
40 | 38 | data_colormap = 'colormapvoc.mat';
|
41 | 39 | case 'cityscapes'
|
42 | 40 | isVal = true;
|
43 |
| - step = 125; |
| 41 | + step = 125; %125=500/4 |
44 | 42 | data_root = '/data2/hszhao/dataset/cityscapes';
|
45 | 43 | eval_list = 'list/cityscapes_val.txt';
|
46 |
| - eval_list = 'list/fine_val.txt'; |
47 | 44 | save_root = 'mc_result/cityscapes/val/pspnet101_713/';
|
48 | 45 | model_weights = 'model/pspnet101_cityscapes.caffemodel';
|
49 | 46 | model_deploy = 'prototxt/pspnet101_cityscapes_713.prototxt';
|
|
55 | 52 | end
|
56 | 53 | skipsize = 0; %skip serveal images in the list
|
57 | 54 |
|
58 |
| -is_save_feat = false; %set to true if final feature map is needed |
| 55 | +is_save_feat = false; %set to true if final feature map is needed (not suggested for storage consuming) |
59 | 56 | save_gray_folder = [save_root 'gray/']; %path for predicted gray image
|
60 | 57 | save_color_folder = [save_root 'color/']; %path for predicted color image
|
61 | 58 | save_feat_folder = [save_root 'feat/']; %path for predicted feature map
|
62 | 59 | scale_array = [1]; %set to [0.5 0.75 1 1.25 1.5 1.75] for multi-scale testing
|
63 |
| -%scale_array = [0.5 0.75 1 1.25 1.5 1.75]; |
64 | 60 | mean_r = 123.68; %means to be subtracted and the given values are used in our training stage
|
65 | 61 | mean_g = 116.779;
|
66 | 62 | mean_b = 103.939;
|
67 | 63 |
|
68 | 64 | acc = double.empty;
|
69 | 65 | iou = double.empty;
|
70 |
| -gpu_id_array = [12:15]; %multi-GPUs for parfor testing, if number of GPUs is changed, remember to change the variable 'step' |
71 |
| -gpu_id_array = [1]; %multi-GPUs for parfor testing, if number of GPUs is changed, remember to change the variable 'step' |
| 66 | +gpu_id_array = [0:3]; %multi-GPUs for parfor testing, if number of GPUs is changed, remember to change the variable 'step' |
72 | 67 | runID = 1;
|
73 | 68 | gpu_num = size(gpu_id_array,2);
|
74 | 69 | index_array = [(runID-1)*gpu_num+1:runID*gpu_num];
|
75 | 70 |
|
76 |
| -for i = 1:gpu_num %change 'parfor' to 'for' if singe GPU testing is used |
| 71 | +parfor i = 1:gpu_num %change 'parfor' to 'for' if singe GPU testing is used |
77 | 72 | eval_sub(data_name,data_root,eval_list,model_weights,model_deploy,fea_cha,base_size,crop_size,data_class,data_colormap, ...
|
78 | 73 | is_save_feat,save_gray_folder,save_color_folder,save_feat_folder,gpu_id_array(i),index_array(i),step,skipsize,scale_array,mean_r,mean_g,mean_b);
|
79 | 74 | end
|
|
0 commit comments