plateform stringclasses 1
value | repo_name stringlengths 13 113 | name stringlengths 3 74 | ext stringclasses 1
value | path stringlengths 12 229 | size int64 23 843k | source_encoding stringclasses 9
values | md5 stringlengths 32 32 | text stringlengths 23 843k |
|---|---|---|---|---|---|---|---|---|
github | qxcv/comp2560-master | variational_descriptor_flow.m | .m | comp2560-master/project/matlab/flow/LDOF_src/src/variational_descriptor_flow.m | 5,432 | utf_8 | 30030984b4b67f377badcaa30f6ca526 | function F = variational_descriptor_flow(im1,im2,...
confidence,centroids1,centroids2,para,verbose)
I1s=getImPyramid(im1,para.sigma,para.downsampling,para.level);
I2s=getImPyramid(im2,para.sigma,para.downsampling,para.level);
gamma = para.gamma;
max_iter = para.out_iter;
downsampling = para.downsampling;
num_lev... |
github | qxcv/comp2560-master | get_pair_dist_matrix.m | .m | comp2560-master/project/matlab/flow/LDOF_src/src/get_pair_dist_matrix.m | 477 | utf_8 | c8d32760ba654393eaf46d5b217b54b6 | function [dist]=get_pair_dist_matrix(E,U_p,V_p)
%number of proposals
L=length(U_p);
e=length(E);
dist=zeros(L,L*e);
for i=1:L
U_pc1=U_p{i};
V_pc1=V_p{i};
for j=i:L
U_pc2=U_p{j};
V_pc2=V_p{j};
dist(i,[j:L:e*L])=penalize(U_pc1(E(:,1)),V_pc1(E(:,1)),...
U_pc2(E(:,2)),V_pc2(... |
github | qxcv/comp2560-master | get_psi_data_weighted.m | .m | comp2560-master/project/matlab/flow/LDOF_src/src/get_psi_data_weighted.m | 269 | utf_8 | 5810e51e5b883e63c1d99cee53745d49 | function psi_data=get_psi_data_weighted(d1square,d2square,gamma,temperature)
w1=1/(1+exp(temperature*(app_sqrt(d1square)-app_sqrt(d2square))));
w2=1-w1;
psi_data=w1.*psiDeriv(d1square)+gamma*w2.*psiDeriv(d2square);
end
function res=app_sqrt(s2)
res=sqrt(s2+eps^2);
end |
github | qxcv/comp2560-master | eval_pix_error.m | .m | comp2560-master/project/matlab/eval/eval_pix_error.m | 2,484 | utf_8 | a378a10ee0f3f7a71449aecb0bf3346c | %
% eval_pix_error: this function returns the average keypoint localization error for
% a given error threshold. accuracy = eval_pix_error(det, gt, thresh) takes a
% pose detection in det, the ground truth in gt, and an error threshold thresh,
% and returns the average accuracy with this threshold for each keypoint.... |
github | qxcv/comp2560-master | evaluate_pose_seqs.m | .m | comp2560-master/project/matlab/eval/evaluate_pose_seqs.m | 687 | utf_8 | cc9226d2596e9899e56c7f2535e8504f | % evaluation of the detected_poses against the ground truth.
function pix_error = evaluate_pose_seqs(detected_pose, gt, pck_thresh)
if nargin == 2
pck_thresh = 15; % default is 15 pix error
end
% first get the annotated poses from the gt corresponding to the images
% used in the detection.
detected_annotated_poses ... |
github | qxcv/comp2560-master | get_framenum.m | .m | comp2560-master/project/matlab/utils/get_framenum.m | 438 | utf_8 | 41a07ea0692d5050e35703369e3b05d5 | % some function to return a unique id for a frame. unique in the dataset.
function frameid = get_framenum(imname)
% this just extracts the number at the end of the filename (e.g.
% 'file-042-00013.png' --> 13)
frame_match = regexp(imname, '[^\d](?<frame_no>\d+)\.png$', 'names');
assert(numel(frame_match) == 1);
str = f... |
github | qxcv/comp2560-master | get_pts_along_limb.m | .m | comp2560-master/project/matlab/utils/get_pts_along_limb.m | 2,702 | utf_8 | 26e4258dbd759d59aafc0eeac3fc35fb | % takes the start and end of limb positions and makes a color histogram descriptor
% for the entire limb rather than patches around the keypoint. numpts
% specifies how many points to be taken from the limb (at equidistant
% points).
function pts = get_pts_along_limb(img, yst, xst, yen, xen, ps, szx, szy, numpts)
regi... |
github | qxcv/comp2560-master | colorhist.m | .m | comp2560-master/project/matlab/utils/colorhist.m | 2,189 | utf_8 | c270c55404ccc1c2f10d624f7d9d70d2 | % assumes the images is color. takes the image and the number of bins in
% each axis. We assume the image is nx2 where the rows represent the pixels
% and the cols are the color channels.
function H = colorhist(img, nbins)
persistent bins;
if isempty(bins)
if isa(img, 'double')% || max(img(:))>1
... |
github | qxcv/comp2560-master | get_rect_coordinates.m | .m | comp2560-master/project/matlab/utils/get_rect_coordinates.m | 1,998 | utf_8 | f10636433c608749bb2445a7da23579b | % returns the coordinates of a rectangle with the main diagonal as the
% coordinates given by start and end.
function region = get_rect_coordinates(startcor, endcor, patch_size, szx, szy)
x1 = startcor(:,2); y1=startcor(:,1); x2 = endcor(:,2); y2=endcor(:,1);
tantheta = (y2-y1)./(x2-x1+eps); theta=atan(tantheta);
w = p... |
github | qxcv/comp2560-master | get_patch_descr_Ex.m | .m | comp2560-master/project/matlab/utils/get_patch_descr_Ex.m | 1,352 | utf_8 | b442f55a20a43c4b229d7d86fa9b3fb7 | function skelclr = get_patch_descr_Ex(skelpts, img, keyjoints, numpts_along_limb, displayposes)
nbins=8;
skelclr = cell(length(keyjoints), numpts_along_limb);
persistent bins;
if isempty(bins)
if max(img(:))<=1 %isa(img, 'double')% ||
bins = linspace(1/nbins,1, nbins);
else
... |
github | qxcv/comp2560-master | get_recombination_tree.m | .m | comp2560-master/project/matlab/utils/get_recombination_tree.m | 1,200 | utf_8 | 0c81d1c85c3522ff04efc3da26c5847c | % returns the joint indices for the recombination tree. Here we combine
% some of the nodes (such as the mid-limbs) returned by Y&R into a single
% limb node when computing the shortest part sequence.
function pose_joints = get_recombination_tree()
neckjoint=1; sholjoint=2; elbowjoint=3; wristjoint=4; % pose tr... |
github | qxcv/comp2560-master | get_motion_from_flow.m | .m | comp2560-master/project/matlab/utils/get_motion_from_flow.m | 2,073 | utf_8 | e6337e841915d8566f540a18052d3338 | function hf= get_motion_from_flow(pt, opticalflow, display_poses)
ps = 15;
u = opticalflow.u; v=opticalflow.v;
n = size(pt,1);
%hf = zeros(n, 2);
str = max(1,pt(:,1)-ps)+1; enr = min(size(u,2), pt(:,1)+ps);
stc = max(1,pt(:,2)-ps)+1; enc = min(size(u,1), pt(:,2)+ps);
% patch_area=zeros(n,1);
% for i=1:n
% patch_ar... |
github | qxcv/comp2560-master | interleave.m | .m | comp2560-master/project/matlab/utils/interleave.m | 320 | utf_8 | da0160861ff958c027c7de8997823688 | % function interleaves to mxn matrices column wise. the result is a
% (mx(2n)) matrix z.
function z = interleave(x,y)
m=size(x,1); n = size(x,2);
if size(x,1)~=size(y,1) || size(x,2)~=size(y,2)
error('incompatible sizes');
end
z = zeros(m,2*n);
for i=1:n
z(:,2*(i-1)+1) = x(:,i);
z(:,2*i) = y(:,i);
end
end
|
github | qxcv/comp2560-master | prepare_dets_for_eval.m | .m | comp2560-master/project/matlab/utils/prepare_dets_for_eval.m | 885 | utf_8 | d42717e20bc342790110df62c291a322 | % prepare for evaluation
function detected_pose_seqs = prepare_dets_for_eval(new_merged_poses, frames, seq)
detected_pose_type = struct('seq', {}, 'filename', {}, 'bestpose',{}, 'all_clear', {});
detected_pose_seqs = repmat(detected_pose_type, [1,1,1]);
mov = 3; d=3; % useful if working with multiple movies and multipl... |
github | qxcv/comp2560-master | get_keypoints.m | .m | comp2560-master/project/matlab/utils/get_keypoints.m | 184 | utf_8 | ba2ab059edd45a86e13e149cc5b23012 | %% get keypoints
function k = get_keypoints(b)
x1 = b(:, 1:4:end);
y1 = b(:, 2:4:end);
x2 = b(:, 3:4:end);
y2 = b(:, 4:4:end);
x = (x1+x2)/2; y = (y1+y2)/2;
k = interleave(x,y);
end%%
|
github | qxcv/comp2560-master | box_normalize.m | .m | comp2560-master/project/matlab/utils/box_normalize.m | 177 | utf_8 | 26cf427d11c14b3948dd36b33a009258 | % normalize the boxes using the image size.
function box = box_normalize(box, imsize)
box(:,1:2:end) = box(:, 1:2:end)/imsize(2);
box(:,2:2:end) = box(:, 2:2:end)/imsize(1);
end |
github | qxcv/comp2560-master | get_patch_descr.m | .m | comp2560-master/project/matlab/utils/get_patch_descr.m | 1,049 | utf_8 | 58fc8ee160e16a569e52eb6501537db1 | % get a patch of size pat_size from img1 around point x,y
function colordesc = get_patch_descr(img, x,y, ps, szx, szy)
N = size(y,1);
%bins = 0:0.01:1; nbins = length(bins);
nbins = 8; colordesc = zeros(512, N);
for i=1:N
try
pat = img(x(i)-ps+1:x(i)+ps, y(i)-ps+1:y(i)+ps,:);
catch
... |
github | qxcv/comp2560-master | mypdist2.m | .m | comp2560-master/project/matlab/utils/mypdist2.m | 5,491 | utf_8 | e54edb4752f416bacb6e521170186462 | % This function belongs to Piotr Dollar's Toolbox
% http://vision.ucsd.edu/~pdollar/toolbox/doc/index.html
% Please refer to the above web page for definitions and clarifications
%
% Calculates the distance between sets of vectors.
%
% Let X be an m-by-p matrix representing m points in p-dimensional space
% and Y be an... |
github | qxcv/comp2560-master | visualizeskeleton.m | .m | comp2560-master/project/matlab/YR/visualization/visualizeskeleton.m | 5,684 | utf_8 | 39e2b8729720405dba8a58449462fea2 | function visualizeskeleton(model)
bs = 4;
% assuming only one component
c = model.components{1};
numparts = length(c);
Nmix = zeros(1,numparts);
for k = 1:numparts
Nmix(k) = length(c(k).filterid);
end
ovec = [0 1 0 -1; 1 0 -1 0];
I = zeros(numparts,size(ovec,2));
for k = 2:numparts
part = c(k);
anchor = zer... |
github | qxcv/comp2560-master | visualizeskeleton.m | .m | comp2560-master/project/matlab/YR/nbest_release/visualizeskeleton.m | 9,373 | utf_8 | e982b2a074418fa08fb43c8fa17a2690 | function visualizeskeleton(model)
bs = 4;
% assuming only one component
c = model.components{1};
numparts = length(c);
Nmix = zeros(1,numparts);
for k = 1:numparts
Nmix(k) = length(c(k).filterid);
end
ovec = [0 1 0 -1; 1 0 -1 0];
I = zeros(numparts,size(ovec,2));
for k = 2:numparts
part = c(k);
anchor =... |
github | qxcv/comp2560-master | visualizemodel.m | .m | comp2560-master/project/matlab/YR/nbest_release/visualizemodel.m | 3,139 | utf_8 | 46088f0e24b8c6e3ef388c4251282ef8 | function visualizemodel(model)
pad = 2;
bs = 20;
% assuming only one component
c = model.components{1};
numparts = length(c);
Nmix = zeros(1,numparts);
for k = 1:numparts
Nmix(k) = length(c(k).filterid);
end
ovec = [0 1 0 -1; 1 0 -1 0];
I = zeros(numparts,size(ovec,2));
for k = 2:numparts
part = c(k);
a... |
github | qxcv/comp2560-master | show_pose_sequence.m | .m | comp2560-master/project/matlab/visualization/show_pose_sequence.m | 1,252 | utf_8 | a39b2a5aef79686019be9ad9803d4d4b | % display the detected pose sequence
function show_pose_sequence(data_path, frames, detected_pose_seq, parent, dest_dir)
if nargin < 5
should_save = false;
else
should_save = true;
mkdir(dest_dir);
end
figno=1023; figure(figno); clf(figno);
for n = 1:length(frames)
video_showskeleton2([data_path fra... |
github | qxcv/comp2560-master | showskeleton.m | .m | comp2560-master/project/matlab/visualization/showskeleton.m | 2,158 | utf_8 | ff9eac60754b610da2907a623f9cc1b3 | % displays a skeleton for the learned piw pose model. use the pts to
% display the poses given keypoints, and use boxes if box coordinates are
% available (e.g., pose boxes returned from loopy_detect_MM.m). np defines
% the limbs to display and visiblility is the occlusion flag.
function showskeleton(img, pts, boxes, ... |
github | qxcv/comp2560-master | imhist_dist.m | .m | comp2560-master/project/matlab/jrcode/imhist_dist.m | 573 | utf_8 | 7500e2e74d77ade6182b5be90f66490f | function dist = imhist_dist(im1, im2, bins_per_channel)
%IMHIST_DIST Compute the cosine similarity between the histograms
%associated with two images.
im1_hist = get_rgb_hist(im1, bins_per_channel);
im2_hist = get_rgb_hist(im2, bins_per_channel);
dist = acos(dot(im1_hist, im2_hist) / (norm(im1_hist) * norm(im2_hist)));... |
github | qxcv/comp2560-master | pcp_new.m | .m | comp2560-master/project/matlab/jrcode/pcp_new.m | 1,620 | utf_8 | 65ebea6e76629df6b4fba5ed7a308242 | function pcps = pcp(preds, gts, limbs)
%PCP Compute Percentage Correct Parts (strict) on input sequences
assert(iscell(preds) && iscell(gts));
assert(length(preds) >= 1);
assert(length(preds) == length(gts));
assert(ismatrix(preds{1}) && size(preds{1}, 2) == 2);
pred_mat = cat(3, preds{:});
gt_mat = cat(3, gts{:});
pc... |
github | qxcv/comp2560-master | get_piw_full.m | .m | comp2560-master/project/matlab/jrcode/get_piw_full.m | 4,582 | utf_8 | 154b934c165fce5ded1a7a629c481077 | function test_seqs = get_piw_full(cache_dir, trans_spec)
%GET_PIW_FULL Get Poses in the Wild dataset
% List of weird gotchas I've found in PIW:
% - Joints which aren't visible (even if their location can be inferred
% despite occlusion) are set to [-1, -1]
% - Joint annotations drift over time, sometimes quite se... |
github | qxcv/comp2560-master | parload.m | .m | comp2560-master/project/matlab/jrcode/parload.m | 358 | utf_8 | c262af4992432d213bb70aa5012a7324 | % par load
function varargout = parload(fullpath, varargin)
varargout = cell(length(varargin), 1);
for ii=1:length(varargin)
var_name = varargin{ii};
assert(isa(class(var_name), 'char'), 'only accept name');
load(fullpath, var_name);
if exist(var_name, 'var')
varargout{ii} = eval(var_name);
else
... |
github | qxcv/comp2560-master | get_mpii_cooking.m | .m | comp2560-master/project/matlab/jrcode/get_mpii_cooking.m | 5,169 | utf_8 | 395502f0ffa62792b7ae8ddc6fd27944 | function test_seqs = get_mpii_cooking(dest_dir, cache_dir, trans_spec)
%GET_MPII_COOKING Fetches continuous pose estimation data from MPII
% train_dataset and val_dataset both come from MPII Cooking's continuous
% pose dataset, with val_dataset scraped from a couple of scenes at the
% end. test_dataset comes from the t... |
github | qxcv/comp2560-master | Compute_SkelFlow.m | .m | comp2560-master/project/matlab/detect/Compute_SkelFlow.m | 7,219 | utf_8 | dea52d3eec5c93995e73f05c59fb1920 | % Perhaps the most important function in this package !
% given a frame-pair, corresponding candidate poses in each frame
% and the optical flow between every consecutive frame pair, this function
% computes the pair-wise cost between every pose pair.
% It takes the the boxes (output of detect_MM) b1 and b2 correspond... |
github | qxcv/comp2560-master | in_frame_detect.m | .m | comp2560-master/project/matlab/detect/in_frame_detect.m | 3,451 | utf_8 | 6043335cc45608efdfd3a8007c659a15 | function boxes = in_frame_detect(count, pyra, unary_map, idpr_map, ...
num_components, components, apps, nms_thresh, nms_parts)
% Detect at most `count` poses in input using a given CNN-computed feature
% pyramid and model.
%
% The function returns a matrix with one row per detected object. The
% last column of ea... |
github | qxcv/comp2560-master | Get_Distance_Weights.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/Get_Distance_Weights.m | 1,725 | utf_8 | 28a98251aadf1bae57947b3aed017b49 | % this function specifies the parameters used in tracking and recombination
% of poses. You might need to play around with these weights over the
% training set to find good combinations. Here are some hints on selecting
% them:
% 1. if there is significant optical flow for a certain body part, then put
% the first tw... |
github | qxcv/comp2560-master | get_groundtruth_for_seq.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/get_groundtruth_for_seq.m | 821 | utf_8 | 595d862b7785b8d7b603155b023a0528 | % extract the ground truth in posedata for the images in the sequence.
% if you use another dataset, you need to change this function accordingly.
function gt = get_groundtruth_for_seq(files, posedata)
persistent gt_framenum;
if isempty(gt_framenum)
gt_framenum = arrayfun(@(x) get_framenum(x.imname), posedata); % eve... |
github | qxcv/comp2560-master | demo.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/demo.m | 4,900 | utf_8 | b7ef5392952e521cad185bad3646f730 | pp% % Copyright (C) 2014 LEAR, Inria Grenoble, France
%
% Permission is hereby granted, free of charge, to any person obtaining
% a copy of this software and associated documentation files (the
% "Software"), to deal in the Software without restriction, including
% without limitation the rights to use, copy, modify, m... |
github | qxcv/comp2560-master | EstimatePosesInVideo.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/EstimatePosesInVideo.m | 9,197 | utf_8 | e43c364912f59303aa5a7dabe75d0856 | function new_merged_poses = EstimatePosesInVideo(src_path, model, ~, config)
global numpts_along_limb num_path_parts;
global parent keyjoints_right keyjoints_left;
global GetDistanceWeightsFn ComputePartPathCostsFn;
%% set some parameters!
flow_param = config.flow_param;
PartIDs = config.PartID... |
github | qxcv/comp2560-master | set_algo_parameters.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/set_algo_parameters.m | 5,891 | utf_8 | 22158733e12f74f06890958bcb744d8f | % set some configuration settings
function config = set_algo_parameters()
%% set the configuration parameters: You need to set this.
% set the cache for storing optical flow and pose candidates. You need a
% large disk space for this.
config.cache_path = './cache/';
%config.cache_path = '/scratch2/bigimbaz/cherian/che... |
github | qxcv/comp2560-master | get_piw_data.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/get_piw_data.m | 6,260 | utf_8 | 5d32c73fdd0f58edfa719d0546920a9d | % read the piw annotations. The annotation database file we use here is
% different in format from the one supplied in the original dataset. This
% code generates the mid-keypoints for the lower and upper-limbs.
%
function pos = get_piw_data(name, piw_seqs_path)
cachedir = './cache/';
cls = [name '_data'];
try
lo... |
github | qxcv/comp2560-master | demo_piw_full.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/demo_piw_full.m | 5,199 | utf_8 | 26f76e460e113507095bae4830daa8ab | % % Copyright (C) 2014 LEAR, Inria Grenoble, France
%
% Permission is hereby granted, free of charge, to any person obtaining
% a copy of this software and associated documentation files (the
% "Software"), to deal in the Software without restriction, including
% without limitation the rights to use, copy, modify, mer... |
github | qxcv/comp2560-master | demo_mpii_cooking.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/demo_mpii_cooking.m | 5,401 | utf_8 | c882ede432b01d007a8c146b9c1b9f06 | % % Copyright (C) 2014 LEAR, Inria Grenoble, France
%
% Permission is hereby granted, free of charge, to any person obtaining
% a copy of this software and associated documentation files (the
% "Software"), to deal in the Software without restriction, including
% without limitation the rights to use, copy, modify, mer... |
github | qxcv/comp2560-master | computeColor.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/flow/LDOF_src/third_party/flow-code-matlab/computeColor.m | 3,142 | utf_8 | a36a650437bc93d4d8ffe079fe712901 | function img = computeColor(u,v)
% computeColor color codes flow field U, V
% According to the c++ source code of Daniel Scharstein
% Contact: schar@middlebury.edu
% Author: Deqing Sun, Department of Computer Science, Brown University
% Contact: dqsun@cs.brown.edu
% $Date: 2007-10-31 21:20:30 (Wed, 31 O... |
github | qxcv/comp2560-master | ann_compile_mex.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/flow/LDOF_src/third_party/ann_mwrapper/ann_compile_mex.m | 1,684 | utf_8 | 33dc804c66a9d68c02efa55511995e33 | function ann_compile_mex()
%ANN_COMPILE_MEX Compiles the core-mex files of the ANN Lib
%
% [ Syntax ]
% - ann_compile_mex
%
% [ Description ]
% - ann_compile_mex re-compiles the mex files.
%
% [ History ]
% - Created by Dahua Lin, on Jul 06, 2007
%
%% configurations
% When you intend to ch... |
github | qxcv/comp2560-master | ann_compile_mex_bak1.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/flow/LDOF_src/third_party/ann_mwrapper/ann_compile_mex_bak1.m | 1,630 | utf_8 | 9a3b317c5db70460d8ea44bd877246d5 | function ann_compile_mex_bak1()
%ANN_COMPILE_MEX Compiles the core-mex files of the ANN Lib
%
% [ Syntax ]
% - ann_compile_mex
%
% [ Description ]
% - ann_compile_mex re-compiles the mex files.
%
% [ History ]
% - Created by Dahua Lin, on Jul 06, 2007
%
%% configurations
% When you intend to change the co... |
github | qxcv/comp2560-master | annquery.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/flow/LDOF_src/third_party/ann_mwrapper/annquery.m | 13,952 | utf_8 | 9ca55bd21175f2b794f2087b4eda6727 | function [nnidx, dists] = annquery(Xr, Xq, k, varargin)
%ANNQUERY Performs Approximate K-Nearest-Neighbor query for a set of points
%
% [ Syntax ]
% - nnidx = annquery(Xr, Xq, k)
% - nnidx = annquery(Xr, Xq, k, ...)
% - [nnidx, dists] = annquery(...)
% - annquery -doc
%
% [ Arguments ]
% - Xr: ... |
github | qxcv/comp2560-master | setopts.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/flow/LDOF_src/third_party/ann_mwrapper/private/setopts.m | 6,161 | utf_8 | e1da88dc99b232199d5082bf6df84068 | function opts = setopts(opts0, varargin)
%SETOPTS Sets the options and makes the option-struct
%
% [ Syntax ]
% - opts = setopts([], name1, value1, name2, value2, ...)
% - opts = setopts([], {name1, value1, name2, value2, ...})
% - opts = setopts([], newopts)
% - opts = setopts(opts0, ...)
%
% [ Argume... |
github | qxcv/comp2560-master | variational_descriptor_flow.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/flow/LDOF_src/src/variational_descriptor_flow.m | 5,432 | utf_8 | 30030984b4b67f377badcaa30f6ca526 | function F = variational_descriptor_flow(im1,im2,...
confidence,centroids1,centroids2,para,verbose)
I1s=getImPyramid(im1,para.sigma,para.downsampling,para.level);
I2s=getImPyramid(im2,para.sigma,para.downsampling,para.level);
gamma = para.gamma;
max_iter = para.out_iter;
downsampling = para.downsampling;
num_lev... |
github | qxcv/comp2560-master | get_pair_dist_matrix.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/flow/LDOF_src/src/get_pair_dist_matrix.m | 477 | utf_8 | c8d32760ba654393eaf46d5b217b54b6 | function [dist]=get_pair_dist_matrix(E,U_p,V_p)
%number of proposals
L=length(U_p);
e=length(E);
dist=zeros(L,L*e);
for i=1:L
U_pc1=U_p{i};
V_pc1=V_p{i};
for j=i:L
U_pc2=U_p{j};
V_pc2=V_p{j};
dist(i,[j:L:e*L])=penalize(U_pc1(E(:,1)),V_pc1(E(:,1)),...
U_pc2(E(:,2)),V_pc2(... |
github | qxcv/comp2560-master | eval_pix_error.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/eval/eval_pix_error.m | 2,487 | utf_8 | d67966324a1f84c6214db69e5b5ba34e | %
% eval_pix_error: this function returns the average keypoint localization error for
% a given error threshold. accuracy = eval_pix_error(det, gt, thresh) takes a
% pose detection in det, the ground truth in gt, and an error threshold thresh,
% and returns the average accuracy with this threshold for each keypoint.... |
github | qxcv/comp2560-master | evaluate_pose_seqs.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/eval/evaluate_pose_seqs.m | 1,262 | utf_8 | 1913c43cfb4362a52deb5fbdeb35cb22 | % evaluation of the detected_poses against the ground truth.
function pix_error = evaluate_pose_seqs(detected_pose, gt, pck_thresh)
if nargin == 2
pck_thresh = 15; % default is 15 pix error
end
% first get the annotated poses from the gt corresponding to the images
% used in the detection.
detected_annotated_poses =... |
github | qxcv/comp2560-master | get_framenum.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/utils/get_framenum.m | 218 | utf_8 | c3be1f6289ed8f89953d225b9b59c0e6 | % some function to return a unique id for a frame. unique in the dataset.
function frameid = get_framenum(imname)
str = strtok(strrep(imname, '-', ''), '.');
str = str(str<'A');
frameid = str2double(str);
end |
github | qxcv/comp2560-master | get_pts_along_limb.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/utils/get_pts_along_limb.m | 2,702 | utf_8 | 26e4258dbd759d59aafc0eeac3fc35fb | % takes the start and end of limb positions and makes a color histogram descriptor
% for the entire limb rather than patches around the keypoint. numpts
% specifies how many points to be taken from the limb (at equidistant
% points).
function pts = get_pts_along_limb(img, yst, xst, yen, xen, ps, szx, szy, numpts)
regi... |
github | qxcv/comp2560-master | colorhist.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/utils/colorhist.m | 2,189 | utf_8 | c270c55404ccc1c2f10d624f7d9d70d2 | % assumes the images is color. takes the image and the number of bins in
% each axis. We assume the image is nx2 where the rows represent the pixels
% and the cols are the color channels.
function H = colorhist(img, nbins)
persistent bins;
if isempty(bins)
if isa(img, 'double')% || max(img(:))>1
... |
github | qxcv/comp2560-master | get_rect_coordinates.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/utils/get_rect_coordinates.m | 1,998 | utf_8 | f10636433c608749bb2445a7da23579b | % returns the coordinates of a rectangle with the main diagonal as the
% coordinates given by start and end.
function region = get_rect_coordinates(startcor, endcor, patch_size, szx, szy)
x1 = startcor(:,2); y1=startcor(:,1); x2 = endcor(:,2); y2=endcor(:,1);
tantheta = (y2-y1)./(x2-x1+eps); theta=atan(tantheta);
w = p... |
github | qxcv/comp2560-master | get_patch_descr_Ex.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/utils/get_patch_descr_Ex.m | 1,352 | utf_8 | b442f55a20a43c4b229d7d86fa9b3fb7 | function skelclr = get_patch_descr_Ex(skelpts, img, keyjoints, numpts_along_limb, displayposes)
nbins=8;
skelclr = cell(length(keyjoints), numpts_along_limb);
persistent bins;
if isempty(bins)
if max(img(:))<=1 %isa(img, 'double')% ||
bins = linspace(1/nbins,1, nbins);
else
... |
github | qxcv/comp2560-master | get_recombination_tree.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/utils/get_recombination_tree.m | 1,171 | utf_8 | a913ae4f2f816146ea61a78a31989f6d | % returns the joint indices for the recombination tree. Here we combine
% some of the nodes (such as the mid-limbs) returned by Y&R into a single
% limb node when computing the shortest part sequence.
function pose_joints = get_recombination_tree()
neckjoint=1; sholjoint=2; elbowjoint=3; wristjoint=4; % pose tr... |
github | qxcv/comp2560-master | get_motion_from_flow.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/utils/get_motion_from_flow.m | 2,149 | utf_8 | fa4cf2c6de39afd8c81dc8c3770aab2f | function hf= get_motion_from_flow(pt, opticalflow, display_poses)
addpath('/home/lear/cherian/projects/pose/MyCode/sequence/shortest-path/');
ps = 15;
u = opticalflow.u; v=opticalflow.v;
n = size(pt,1);
%hf = zeros(n, 2);
str = max(1,pt(:,1)-ps)+1; enr = min(size(u,2), pt(:,1)+ps);
stc = max(1,pt(:,2)-ps)+1; enc = min... |
github | qxcv/comp2560-master | interleave.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/utils/interleave.m | 320 | utf_8 | da0160861ff958c027c7de8997823688 | % function interleaves to mxn matrices column wise. the result is a
% (mx(2n)) matrix z.
function z = interleave(x,y)
m=size(x,1); n = size(x,2);
if size(x,1)~=size(y,1) || size(x,2)~=size(y,2)
error('incompatible sizes');
end
z = zeros(m,2*n);
for i=1:n
z(:,2*(i-1)+1) = x(:,i);
z(:,2*i) = y(:,i);
end
end
|
github | qxcv/comp2560-master | prepare_dets_for_eval.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/utils/prepare_dets_for_eval.m | 885 | utf_8 | d42717e20bc342790110df62c291a322 | % prepare for evaluation
function detected_pose_seqs = prepare_dets_for_eval(new_merged_poses, frames, seq)
detected_pose_type = struct('seq', {}, 'filename', {}, 'bestpose',{}, 'all_clear', {});
detected_pose_seqs = repmat(detected_pose_type, [1,1,1]);
mov = 3; d=3; % useful if working with multiple movies and multipl... |
github | qxcv/comp2560-master | get_keypoints.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/utils/get_keypoints.m | 184 | utf_8 | ba2ab059edd45a86e13e149cc5b23012 | %% get keypoints
function k = get_keypoints(b)
x1 = b(:, 1:4:end);
y1 = b(:, 2:4:end);
x2 = b(:, 3:4:end);
y2 = b(:, 4:4:end);
x = (x1+x2)/2; y = (y1+y2)/2;
k = interleave(x,y);
end%%
|
github | qxcv/comp2560-master | box_normalize.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/utils/box_normalize.m | 177 | utf_8 | 26cf427d11c14b3948dd36b33a009258 | % normalize the boxes using the image size.
function box = box_normalize(box, imsize)
box(:,1:2:end) = box(:, 1:2:end)/imsize(2);
box(:,2:2:end) = box(:, 2:2:end)/imsize(1);
end |
github | qxcv/comp2560-master | get_patch_descr.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/utils/get_patch_descr.m | 1,049 | utf_8 | 58fc8ee160e16a569e52eb6501537db1 | % get a patch of size pat_size from img1 around point x,y
function colordesc = get_patch_descr(img, x,y, ps, szx, szy)
N = size(y,1);
%bins = 0:0.01:1; nbins = length(bins);
nbins = 8; colordesc = zeros(512, N);
for i=1:N
try
pat = img(x(i)-ps+1:x(i)+ps, y(i)-ps+1:y(i)+ps,:);
catch
... |
github | qxcv/comp2560-master | mypdist2.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/utils/mypdist2.m | 5,491 | utf_8 | e54edb4752f416bacb6e521170186462 | % This function belongs to Piotr Dollar's Toolbox
% http://vision.ucsd.edu/~pdollar/toolbox/doc/index.html
% Please refer to the above web page for definitions and clarifications
%
% Calculates the distance between sets of vectors.
%
% Let X be an m-by-p matrix representing m points in p-dimensional space
% and Y be an... |
github | qxcv/comp2560-master | visualizeskeleton.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/YR/visualization/visualizeskeleton.m | 5,684 | utf_8 | 39e2b8729720405dba8a58449462fea2 | function visualizeskeleton(model)
bs = 4;
% assuming only one component
c = model.components{1};
numparts = length(c);
Nmix = zeros(1,numparts);
for k = 1:numparts
Nmix(k) = length(c(k).filterid);
end
ovec = [0 1 0 -1; 1 0 -1 0];
I = zeros(numparts,size(ovec,2));
for k = 2:numparts
part = c(k);
anchor = zer... |
github | qxcv/comp2560-master | visualizeskeleton.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/YR/nbest_release/visualizeskeleton.m | 9,373 | utf_8 | e982b2a074418fa08fb43c8fa17a2690 | function visualizeskeleton(model)
bs = 4;
% assuming only one component
c = model.components{1};
numparts = length(c);
Nmix = zeros(1,numparts);
for k = 1:numparts
Nmix(k) = length(c(k).filterid);
end
ovec = [0 1 0 -1; 1 0 -1 0];
I = zeros(numparts,size(ovec,2));
for k = 2:numparts
part = c(k);
anchor =... |
github | qxcv/comp2560-master | detect_MM.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/YR/nbest_release/detect_MM.m | 7,903 | utf_8 | 939c994f2fd0ec942c4c1f69d150d93f | function boxes = detect_MM(im, model,thresh,partIDs)
% This algorithm approximates nbest maximal decoders introduced in the paper,
% returning high-scoring and diverse poses.
% Greedy NMS is applied to max-marginal score map for each part.
% The poses are obtained by backtracking from the location returned by NMS.
% ... |
github | qxcv/comp2560-master | visualizemodel.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/YR/nbest_release/visualizemodel.m | 3,139 | utf_8 | 46088f0e24b8c6e3ef388c4251282ef8 | function visualizemodel(model)
pad = 2;
bs = 20;
% assuming only one component
c = model.components{1};
numparts = length(c);
Nmix = zeros(1,numparts);
for k = 1:numparts
Nmix(k) = length(c(k).filterid);
end
ovec = [0 1 0 -1; 1 0 -1 0];
I = zeros(numparts,size(ovec,2));
for k = 2:numparts
part = c(k);
a... |
github | qxcv/comp2560-master | show_pose_sequence.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/visualization/show_pose_sequence.m | 1,021 | utf_8 | cfb24a2877c9e81b910fb1a89729ebb3 | % display the detected pose sequence
function show_pose_sequence(data_path, frames, detected_pose_seq)
figno=1023; figure(figno);clf(figno);
for n = 1:length(frames)
%video_showskeleton2(img{n}, detected_pose_seq(n,:));axis off;
video_showskeleton2([data_path frames(n).name], detected_pose_seq(n,:));axis o... |
github | qxcv/comp2560-master | showskeleton.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/visualization/showskeleton.m | 2,158 | utf_8 | ff9eac60754b610da2907a623f9cc1b3 | % displays a skeleton for the learned piw pose model. use the pts to
% display the poses given keypoints, and use boxes if box coordinates are
% available (e.g., pose boxes returned from loopy_detect_MM.m). np defines
% the limbs to display and visiblility is the occlusion flag.
function showskeleton(img, pts, boxes, ... |
github | qxcv/comp2560-master | imhist_dist.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/jrcode/imhist_dist.m | 573 | utf_8 | 7500e2e74d77ade6182b5be90f66490f | function dist = imhist_dist(im1, im2, bins_per_channel)
%IMHIST_DIST Compute the cosine similarity between the histograms
%associated with two images.
im1_hist = get_rgb_hist(im1, bins_per_channel);
im2_hist = get_rgb_hist(im2, bins_per_channel);
dist = acos(dot(im1_hist, im2_hist) / (norm(im1_hist) * norm(im2_hist)));... |
github | qxcv/comp2560-master | get_piw_full.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/jrcode/get_piw_full.m | 4,582 | utf_8 | 154b934c165fce5ded1a7a629c481077 | function test_seqs = get_piw_full(cache_dir, trans_spec)
%GET_PIW_FULL Get Poses in the Wild dataset
% List of weird gotchas I've found in PIW:
% - Joints which aren't visible (even if their location can be inferred
% despite occlusion) are set to [-1, -1]
% - Joint annotations drift over time, sometimes quite se... |
github | qxcv/comp2560-master | parload.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/jrcode/parload.m | 358 | utf_8 | c262af4992432d213bb70aa5012a7324 | % par load
function varargout = parload(fullpath, varargin)
varargout = cell(length(varargin), 1);
for ii=1:length(varargin)
var_name = varargin{ii};
assert(isa(class(var_name), 'char'), 'only accept name');
load(fullpath, var_name);
if exist(var_name, 'var')
varargout{ii} = eval(var_name);
else
... |
github | qxcv/comp2560-master | get_mpii_cooking.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/jrcode/get_mpii_cooking.m | 5,169 | utf_8 | 395502f0ffa62792b7ae8ddc6fd27944 | function test_seqs = get_mpii_cooking(dest_dir, cache_dir, trans_spec)
%GET_MPII_COOKING Fetches continuous pose estimation data from MPII
% train_dataset and val_dataset both come from MPII Cooking's continuous
% pose dataset, with val_dataset scraped from a couple of scenes at the
% end. test_dataset comes from the t... |
github | qxcv/comp2560-master | Compute_SkelFlow.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/detect/Compute_SkelFlow.m | 6,990 | utf_8 | 4555f8c3d917f003303ae6ca39eb6a4b | % Perhaps the most important function in this package !
% given a frame-pair, corresponding candidate poses in each frame
% and the optical flow between every consecutive frame pair, this function
% computes the pair-wise cost between every pose pair.
% It takes the the boxes (output of detect_MM) b1 and b2 correspond... |
github | qxcv/comp2560-master | flowpyramid.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/detect/flowpyramid.m | 3,177 | utf_8 | 57abbc1f2b62924c245cfec574dbd205 | function flowpyra = flowpyramid(img1, img2, fim, model)
% Compute flow pyramid.
%
% pyra.feat{i} is the i-th level of the feature pyramid.
% pyra.scales{i} is the scaling factor used for the i-th level.
% pyra.feat{i+interval} is computed at exactly half the resolution of feat{i}.
% first octave halucinates higher reso... |
github | qxcv/comp2560-master | loopy_detect_MM.m | .m | comp2560-master/thirdparty/cherian-et-al-cvpr-14/detect/loopy_detect_MM.m | 17,067 | utf_8 | 6c6fba12654f8a5ee2b54423ab604918 | function boxes = loopy_detect_MM(im1, im2, fim, model, thresh, cycled_nodes, PartIDs, flow_param)
% boxes = detect(im1, im2, fim, model, thresh)
% also uses optical flow between im1 and im2 into the detection of poses in
% consecutive frames simultaneously.
% Detect objects in input using a model and a score threshold.... |
github | qxcv/comp2560-master | global_conf.m | .m | comp2560-master/thirdparty/chen-et-al-nips-14/global_conf.m | 1,492 | utf_8 | c62ad7fbaabd3f8cd10a7d9358fb7bd0 | function conf = global_conf()
assert_not_in_parallel_worker();
% dataset
conf.interval = 10; % 10 levels from 1 to 1/2
conf.memsize = 0.5; % 0.5 gb
conf.NEG_N = 80;
conf.device_id = 0;
conf.caffe_root = './external/caffe';
% default configurations
conf.mining_neg = true;
conf.mining_pos = false;
conf.K = 13;
conf.test_... |
github | qxcv/comp2560-master | imreadx.m | .m | comp2560-master/thirdparty/chen-et-al-nips-14/dataio/imreadx.m | 402 | utf_8 | a7ae5a0fcb499f30e8f8fba7e9188a39 | function im = imreadx(ex)
im = make_color(imread(ex.im));
% !!! always flip first !!!!
if ex.isflip
im = im(:,end:-1:1,:);
end
% rotate and flip
if ex.r_degree ~= 0
im = imrotate(im, ex.r_degree);
end
function im = make_color(input)
% Convert input image to color.
% im = color(input)
if size(input, 3) == 1
im... |
github | qxcv/comp2560-master | lookupPart.m | .m | comp2560-master/thirdparty/chen-et-al-nips-14/tools/lookupPart.m | 2,630 | utf_8 | b2f0ea899418570ee5125f8eb57ff2a1 | function [inds numParts] = lookupPart(varargin)
% Returns the part # in the annotation for a string representation
key = get_key();
invkey = fields(key);
assert(key.(invkey{10})==10);
if nargout==2
numParts = max(cellfun(@(x)key.(x), fieldnames(rmfield(key,'KEYPOINT_FLIPMAP'))));
end
if nargin == 0,
inds =... |
github | qxcv/comp2560-master | testoverlap.m | .m | comp2560-master/thirdparty/chen-et-al-nips-14/tools/testoverlap.m | 1,101 | utf_8 | e8a8b1b4a54dbc39dbe95e6da7b21fa4 | % Compute a mask of filter reponse locations (for a filter of size sizy,sizx)
% that sufficiently overlap a ground-truth bounding box (bbox)
% at a particular level in a feature pyramid
function ov = testoverlap(sizx,sizy,ud1,ud2,pyra,bbox,overlap)
scale = pyra.scale;
% ---------- TODO -----------
padx = pyra.padx;
pa... |
github | qxcv/comp2560-master | parload.m | .m | comp2560-master/thirdparty/chen-et-al-nips-14/tools/parload.m | 358 | utf_8 | c262af4992432d213bb70aa5012a7324 | % par load
function varargout = parload(fullpath, varargin)
varargout = cell(length(varargin), 1);
for ii=1:length(varargin)
var_name = varargin{ii};
assert(isa(class(var_name), 'char'), 'only accept name');
load(fullpath, var_name);
if exist(var_name, 'var')
varargout{ii} = eval(var_name);
else
... |
github | qxcv/comp2560-master | parsave.m | .m | comp2560-master/thirdparty/chen-et-al-nips-14/tools/parsave.m | 254 | utf_8 | 8216068114fa720295fb4518b561ca44 | % par save
function parsave(fname,data, varargin)
if ~isempty(varargin)
var_name = varargin{1};
else
var_name=genvarname(inputname(2));
end
eval([var_name '=data;']);
try
save(fname,var_name,'-append')
catch
save(fname,var_name)
end
|
github | qxcv/comp2560-master | crop_patch.m | .m | comp2560-master/thirdparty/chen-et-al-nips-14/tools/crop_patch.m | 2,409 | utf_8 | 52820e3eb681c58ec5f40dec5239f372 | function cpatch = crop_patch(imdata, label, psize)
persistent NEG_N;
if isempty(NEG_N)
conf = global_conf();
NEG_N = conf.NEG_N;
end
RNG_MAX = 1e9 - 1;
is_negative = isempty(imdata.joints);
if ~is_negative
% negative images
joints = imdata.joints;
scale_x = imdata.scale_x;
scale_y = imdata.scale_y;
gla... |
github | qxcv/comp2560-master | lookupPart.m | .m | comp2560-master/thirdparty/chen-et-al-nips-14/dataset/FLIC/lookupPart.m | 2,630 | utf_8 | b2f0ea899418570ee5125f8eb57ff2a1 | function [inds numParts] = lookupPart(varargin)
% Returns the part # in the annotation for a string representation
key = get_key();
invkey = fields(key);
assert(key.(invkey{10})==10);
if nargout==2
numParts = max(cellfun(@(x)key.(x), fieldnames(rmfield(key,'KEYPOINT_FLIPMAP'))));
end
if nargin == 0,
inds =... |
github | qxcv/comp2560-master | qp_write.m | .m | comp2560-master/thirdparty/chen-et-al-nips-14/external/qpsolver/qp_write.m | 1,505 | utf_8 | 229c74c3ca521f58c83a77c174e6d3f6 | % qp_write(ex)
% where ex.id = K X 1
% ex.blocks(j).i = starting index of j^th feature block
% ex.blocks(j).x = feature block
% final feature = [ex.blocks(:).x]
%
% Write ex to a QP of the form:
% min_{w,e} ||(w-w0)*r||^2 + sum_i c_i e_i
% s.t. w x_ij >= 1 - e_i
%
% We can write the above Q... |
github | qxcv/comp2560-master | qp_opt.m | .m | comp2560-master/thirdparty/chen-et-al-nips-14/external/qpsolver/qp_opt.m | 2,028 | utf_8 | 02ab5a783b9007f2334c987d6458579b | function qp_opt(tol,iter)
% qp_opt(tol,iter)
% Optimize QP until relative difference between lower and upper bound is below 'tol'
global qp;
if nargin < 1,
tol = .05;
end
if nargin < 2,
iter = 3000;
end
% Recompute qp.w in case of numerical precision issues
qp_refresh();
C = 1;
I = 1:qp.n;
[id,J] = sortrows(qp... |
github | qxcv/comp2560-master | qp_prune.m | .m | comp2560-master/thirdparty/chen-et-al-nips-14/external/qpsolver/qp_prune.m | 828 | utf_8 | 1789b75e3aa75f855dffc94972d374c5 | % Prunes qp to current active constraints (eg., support vectors)
function n = qp_prune
global qp
% if cache is full of support vectors, only keep non-zero (and fixed) ones
if all(qp.sv),
qp.sv = qp.a > 0;
qp.sv(qp.svfix) = 1;
end
I = find(qp.sv > 0);
n = length(I);
assert(n > 0);
qp.l = 0;
qp.w = zeros(size(qp.w... |
github | qxcv/comp2560-master | qp_one.m | .m | comp2560-master/thirdparty/chen-et-al-nips-14/external/qpsolver/qp_one.m | 4,224 | utf_8 | c0f146c59f6bbdbb3be16decf2ff57e0 | % Perform one pass through current set of support vectors
function qp_one
global qp
MEX = true;
%MEX = false;
% Random ordering of support vectors
I = find(qp.sv);
I = I(randperm(length(I)));
assert(~isempty(I));
% Mex file is much faster
if MEX,
loss = qp_one_sparse(qp.x,qp.i,qp.b,qp.d,qp.... |
github | qxcv/comp2560-master | detect.m | .m | comp2560-master/thirdparty/chen-et-al-nips-14/src/detect.m | 8,000 | utf_8 | 33da8db79e0106ff09309d4d3b7f1aa7 | function [boxes,model,ex] = detect(iminfo, model, thresh, bbox, overlap, id, label)
% Detect objects in image using a model and a score threshold.
% Higher threshold leads to fewer detections.
%
% The function returns a matrix with one row per detected object. The
% last column of each row gives the score of the detec... |
github | qxcv/comp2560-master | detect_fast.m | .m | comp2560-master/thirdparty/chen-et-al-nips-14/src/detect_fast.m | 4,913 | utf_8 | 308f6305f946a050cd5b8a5ca166e127 | function boxes = detect_fast(iminfo, model, thresh, param)
% boxes = detect(im, model, thresh)
% Detect objects in input using a model and a score threshold.
% Higher threshold leads to fewer detections.
%
% The function returns a matrix with one row per detected object. The
% last column of each row gives the score o... |
github | qxcv/comp2560-master | assign_label.m | .m | comp2560-master/thirdparty/chen-et-al-nips-14/src/assign_label.m | 1,931 | utf_8 | 00ee236c8016d93f03a808c47efd9fb0 | function labels = assign_label(imdata, clusters, pa, tsize, K, is_check)
if ~exist('is_check', 'var')
is_check = false;
end
% add mix field to imgs
p_no = numel(pa);
labels = struct( 'mix_id', cell(numel(imdata), 1), ...
'global_id', cell(numel(imdata), 1), ...
'near', cell(numel(imdata), 1), ...
'invalid', cel... |
github | qxcv/comp2560-master | train.m | .m | comp2560-master/thirdparty/chen-et-al-nips-14/src/train.m | 6,264 | utf_8 | 03456f1691ac1b66e99a8ffb3d1d6f0b | function model = train(name, model, pos, neg, iter, C, wpos, maxsize, overlap)
% Train a structured SVM with latent assignement of positive variables
% pos = list of positive images with part annotations
% neg = list of negative images
% iter is the number of training iterations
% C = scale factor for slack loss
%... |
github | qxcv/comp2560-master | defvector.m | .m | comp2560-master/thirdparty/chen-et-al-nips-14/src/defvector.m | 393 | utf_8 | 613231d473f11ad8cd85fa2524f18600 | % Compute the deformation feature given parent locations,
% child locations, and the child part
function res = defvector(part, x1, y1, x2, y2, m, id)
probx = x1 - part.mean_x{id}(m);
proby = y1 - part.mean_y{id}(m);
var_x = part.var_x{id}(m);
var_y = part.var_y{id}(m);
dx = (probx - x2) / var_x;
dy = (proby - y2) / v... |
github | qxcv/comp2560-master | modelcomponents.m | .m | comp2560-master/thirdparty/chen-et-al-nips-14/src/modelcomponents.m | 1,407 | utf_8 | 0e43dd3ddd208b268362d16602cbd988 | % Cache various statistics from the model data structure for later use
function [components,apps] = modelcomponents(model)
components = cell(length(model.components),1);
for c = 1:length(model.components)
for k = 1:length(model.components{c})
p = model.components{c}(k); % has nbh_IDs
nbh_N = numel(p.nbh_IDs);... |
github | qxcv/comp2560-master | detect_fast.m | .m | comp2560-master/thirdparty/yang-ramanan-2011/code-basic/detect_fast.m | 6,012 | utf_8 | 8a6fb2a4be5cce75d990142f69756aae | function boxes = detect_fast(im, model)
% boxes = detect(im, model, thresh)
% Detect objects in input using a model and a score threshold.
% Higher threshold leads to fewer detections.
%
% The function returns a matrix with one row per detected object. The
% last column of each row gives the score of the detection. T... |
github | qxcv/comp2560-master | visualizeskeleton.m | .m | comp2560-master/thirdparty/yang-ramanan-2011/code-basic/visualization/visualizeskeleton.m | 5,684 | utf_8 | 39e2b8729720405dba8a58449462fea2 | function visualizeskeleton(model)
bs = 4;
% assuming only one component
c = model.components{1};
numparts = length(c);
Nmix = zeros(1,numparts);
for k = 1:numparts
Nmix(k) = length(c(k).filterid);
end
ovec = [0 1 0 -1; 1 0 -1 0];
I = zeros(numparts,size(ovec,2));
for k = 2:numparts
part = c(k);
anchor = zer... |
github | qxcv/comp2560-master | detect.m | .m | comp2560-master/thirdparty/yang-ramanan-2011/code-full/detection/detect.m | 10,674 | utf_8 | 3c4d0e6e81fe35cb91a852c0099ef710 | function [boxes,model,ex] = detect(im, model, thresh, bbox, overlap, id, label)
% [boxes,model,ex] = detect(im, model, thresh, bbox, overlap, id, label)
% Detect objects in image using a model and a score threshold.
% Higher threshold leads to fewer detections.
%
% The function returns a matrix with one row per ... |
github | qxcv/comp2560-master | detect_fast.m | .m | comp2560-master/thirdparty/yang-ramanan-2011/code-full/detection/detect_fast.m | 5,762 | utf_8 | 5431046a2df998e60a41d767e1948be7 | function boxes = detect_fast(im, model, thresh)
% boxes = detect(im, model, thresh)
% Detect objects in input using a model and a score threshold.
% Higher threshold leads to fewer detections.
%
% The function returns a matrix with one row per detected object. The
% last column of each row gives the score of the detec... |
github | qxcv/comp2560-master | qp_write.m | .m | comp2560-master/thirdparty/yang-ramanan-2011/code-full/learning/qp_write.m | 1,505 | utf_8 | 229c74c3ca521f58c83a77c174e6d3f6 | % qp_write(ex)
% where ex.id = K X 1
% ex.blocks(j).i = starting index of j^th feature block
% ex.blocks(j).x = feature block
% final feature = [ex.blocks(:).x]
%
% Write ex to a QP of the form:
% min_{w,e} ||(w-w0)*r||^2 + sum_i c_i e_i
% s.t. w x_ij >= 1 - e_i
%
% We can write the above Q... |
github | qxcv/comp2560-master | qp_opt.m | .m | comp2560-master/thirdparty/yang-ramanan-2011/code-full/learning/qp_opt.m | 2,007 | utf_8 | cc1ea2033343139df2642d05930a57b7 | function qp_opt(tol,iter)
% qp_opt(tol,iter)
% Optimize QP until relative difference between lower and upper bound is below 'tol'
global qp;
if nargin < 1,
tol = .05;
end
if nargin < 2,
iter = 1000;
end
% Recompute qp.w in case of numerical precision issues
qp_refresh();
C = 1;
I = 1:qp.n;
[id,J] = sortrows(qp... |
github | qxcv/comp2560-master | qp_prune.m | .m | comp2560-master/thirdparty/yang-ramanan-2011/code-full/learning/qp_prune.m | 828 | utf_8 | 1789b75e3aa75f855dffc94972d374c5 | % Prunes qp to current active constraints (eg., support vectors)
function n = qp_prune
global qp
% if cache is full of support vectors, only keep non-zero (and fixed) ones
if all(qp.sv),
qp.sv = qp.a > 0;
qp.sv(qp.svfix) = 1;
end
I = find(qp.sv > 0);
n = length(I);
assert(n > 0);
qp.l = 0;
qp.w = zeros(size(qp.w... |
github | qxcv/comp2560-master | qp_one.m | .m | comp2560-master/thirdparty/yang-ramanan-2011/code-full/learning/qp_one.m | 4,224 | utf_8 | c0f146c59f6bbdbb3be16decf2ff57e0 | % Perform one pass through current set of support vectors
function qp_one
global qp
MEX = true;
%MEX = false;
% Random ordering of support vectors
I = find(qp.sv);
I = I(randperm(length(I)));
assert(~isempty(I));
% Mex file is much faster
if MEX,
loss = qp_one_sparse(qp.x,qp.i,qp.b,qp.d,qp.... |
github | qxcv/comp2560-master | train.m | .m | comp2560-master/thirdparty/yang-ramanan-2011/code-full/learning/train.m | 6,358 | utf_8 | c2623dfaa490856f108563245b71ebf1 | function model = train(name, model, pos, neg, warp, iter, C, wpos, maxsize, overlap)
% model = train(name, model, pos, neg, warp, iter, C, Jpos, maxsize, overlap)
% 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
% Train a structured SVM with latent assignement of positive variables
% pos = list... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.