mirror of
https://gitcode.com/gh_mirrors/ope/OpenFace.git
synced 2026-05-16 20:28:00 +00:00
123 lines
4.3 KiB
Matlab
123 lines
4.3 KiB
Matlab
clear;
|
|
addpath(genpath('../'));
|
|
|
|
output_dir = './CLM_res/';
|
|
|
|
%% select database and load bb initializations
|
|
db_root = 'C:\datasets\300VW_Dataset_2015_12_14/';
|
|
bb_root = './300VW_dets_mtcnn/';
|
|
extra_dir = 'C:\datasets\300VW_Dataset_2015_12_14\extra';
|
|
[ vid_locs, bboxes, gts_all, invalid_frames ] = CollectTestData(db_root, bb_root, extra_dir);
|
|
|
|
%% loading the patch experts and the PDM
|
|
[ patches, pdm, clmParams ] = Load_CLM_general();
|
|
|
|
% Adapt the parameters for video
|
|
clmParams.window_size = [25,25; 23,23; 21,21; 19,19; 17,17;];
|
|
clmParams.numPatchIters = size(clmParams.window_size,1);
|
|
clmParams.regFactor = [35, 27, 20, 20];
|
|
clmParams.sigmaMeanShift = [1.25, 1.375, 1.5, 1.5];
|
|
clmParams.tikhonov_factor = [0,0,0, 0];
|
|
clmParams.numPatchIters = 4;
|
|
|
|
num_points = numel(pdm.M) / 3;
|
|
|
|
multi_view = true;
|
|
verbose = true;
|
|
|
|
%% Select video
|
|
for i=1:numel(vid_locs)
|
|
|
|
vid = VideoReader(vid_locs{i});
|
|
|
|
bounding_boxes = bboxes{i};
|
|
preds = [];
|
|
n_frames = size(bounding_boxes,1);
|
|
for f=1:n_frames
|
|
input_image = readFrame(vid);
|
|
|
|
%% Initialize from detected bounding boxes every 30 frames
|
|
if (mod(f-1, 30) == 0)
|
|
ind = min(f, size(bounding_boxes,1));
|
|
bb = bounding_boxes(ind, :);
|
|
% If no face detected use the closest detected BB
|
|
if(bb(3) == 0)
|
|
ind_next = ind + find(bounding_boxes(ind+1:end,3), 1);
|
|
if(isempty(bounding_boxes(ind_next)) || bounding_boxes(ind_next,3)==0)
|
|
ind_next = find(bounding_boxes(1:ind,3), 1, 'last');
|
|
end
|
|
bb = bounding_boxes(ind_next, :);
|
|
end
|
|
bb(3) = bb(1) + bb(3);
|
|
bb(4) = bb(2) + bb(4);
|
|
reset = true;
|
|
else
|
|
reset = false;
|
|
end
|
|
|
|
% have a multi-view version for initialization, otherwise use
|
|
% previous shape
|
|
if(reset && multi_view)
|
|
clmParams.window_size = [25,25; 23,23; 21,21; 21,21];
|
|
clmParams.numPatchIters = 4;
|
|
clmParams.startScale = 1;
|
|
|
|
% The number of hyps is different in CLM from CE-CLM as too
|
|
% many actually harm it
|
|
views = [0,0,0; 0,-30,0; 0,30,0; 0,-55,0; 0,55,0; 0,0,30; 0,0,-30];
|
|
views = views * pi/180;
|
|
|
|
shapes = zeros(num_points, 2, size(views,1));
|
|
ls = zeros(size(views,1),1);
|
|
lmark_lhoods = zeros(num_points,size(views,1));
|
|
views_used = zeros(num_points,size(views,1));
|
|
g_params = cell(size(views,1),1);
|
|
l_params = cell(size(views,1),1);
|
|
|
|
% Find the best orientation
|
|
for v = 1:size(views,1)
|
|
[shapes(:,:,v),g_params{v},l_params{v},ls(v),lmark_lhoods(:,v),views_used(v)] = Fitting_from_bb(input_image, [], bb, pdm, patches, clmParams, 'orientation', views(v,:));
|
|
end
|
|
|
|
[lhood, v_ind] = max(ls);
|
|
lmark_lhood = lmark_lhoods(:,v_ind);
|
|
g_param = g_params{v_ind};
|
|
l_param = l_params{v_ind};
|
|
shape = shapes(:,:,v_ind);
|
|
view_used = views_used(v);
|
|
|
|
else
|
|
clmParams.window_size = [23,23; 21,21; 19,19; 17,17];
|
|
clmParams.numPatchIters = 3;
|
|
clmParams.startScale = 2;
|
|
[shape,g_param,l_param,lhood,lmark_lhood,view_used] = Fitting_from_bb(input_image, [], bb, pdm, patches, clmParams, 'gparam', g_param, 'lparam', l_param);
|
|
end
|
|
|
|
preds = cat(3, preds, shape);
|
|
|
|
%% plot the result
|
|
imshow(input_image);
|
|
hold on;
|
|
plot(shape(:,1), shape(:,2), '.r');
|
|
% rectangle('Position', [bb(2), bb(1), bb(4), bb(3)]);
|
|
hold off;
|
|
drawnow expose
|
|
|
|
end
|
|
|
|
%% Grab the ground truth
|
|
gt_landmarks = gts_all{i};
|
|
|
|
% Remove unreliable frames
|
|
if(~isempty(invalid_frames{i}))
|
|
preds(:,:,int32(invalid_frames{i}))=[];
|
|
end
|
|
|
|
if(size(gt_landmarks,3) ~= size(preds,3))
|
|
fprintf('something went wrong with vid %d\n', i);
|
|
end
|
|
|
|
[vid_name,~,~] = fileparts(vid_locs{i});
|
|
[~,vid_name,~] = fileparts(vid_name);
|
|
save([output_dir, '/', vid_name], 'preds', 'gt_landmarks');
|
|
end |