This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Psuedo Code for Multi-Perspective Data Validation

Algorithm 1 Validate Multi-Perspective Data
1:Input:
2:𝕍\mathbb{V}: List of valid camera views
3:Lv{L}_{\textit{v}}: Valid 2D keypoints for each view v∈𝕍\textit{v}\in\mathbb{V}.
4:Kv,Mv{K}_{\textit{v}},{M}_{\textit{v}}: Camera and transformation matrices for each view v∈𝕍\textit{v}\in\mathbb{V}.
5:Jv{J}_{\textit{v}}: 3D joint pose computed through keypoint optimization of multiple view excluding view v∈𝕍\textit{v}\in\mathbb{V},
6:Output:
7:𝕆\mathbb{O}: List of outlier views.
8:𝕀\mathbb{I}: List of inlier views.
9:procedure ValidateMP(Lv,Kv,Mv,Jv{L}_{\textit{v}},{K}_{\textit{v}},{M}_{\textit{v}},{J}_{\textit{v}})
10:    𝕆←l​i​s​t​()\mathbb{O}\leftarrow list()
11:    𝕀←l​i​s​t​()\mathbb{I}\leftarrow list()
12:    for each valid view tt in 𝕍\mathbb{V} do
13:       for each view v′v^{\prime} in 𝕍\mathbb{V} do
14:          Transform 3D joints to camera coords and projection to 2D:
15:          c​a​m​Kv′=Kv′⋅Mv′⋅Jt\quad camK_{v^{\prime}}=K_{v^{\prime}}\cdot M_{v^{\prime}}\cdot J_{t}
16:          Compute MSE with valid keypoints:
17:          d​i​f​fv′=M​S​E​(Lv′,c​a​m​Kv′)\quad{diff}_{v^{\prime}}=MSE(L_{v^{\prime}},camK_{v^{\prime}})
18:       end for
19:       Calculate not optimized MSE for view tt:
20:       d​i​f​fn​o​t​_​o​p​t​m,t=d​i​f​ft\quad{diff}_{not\_optm,t}={diff}_{t}
21:       Calculate average optimized MSE excluding view tt:
22:       d​i​f​fo​p​t​m,t=1|V|−1​∑v∈V,v≠td​i​f​fv\quad{diff}_{optm,t}=\frac{1}{|V|-1}\sum_{v\in V,v\neq t}{diff}_{v}
23:    end for
24:    if ∀t,d​i​f​fn​o​t​_​o​p​t​m,t≥t​h​r​e​s​h​o​l​d\forall t,{diff}_{not\_optm,t}\geq threshold then
25:       Mark all views tt as outliers.
26:    else
27:       tm​a​x=arg⁡maxt⁡d​i​f​fn​o​t​_​o​p​t​m,tt_{max}=\arg\max_{t}{diff}_{not\_optm,t}
28:       tm​i​n=arg⁡mint⁡d​i​f​fo​p​t​m,tt_{min}=\arg\min_{t}{diff}_{optm,t}
29:       for t in 𝕍\mathbb{V} do:
30:          if t=tm​a​xt=t_{max} and d​i​f​fn​o​t​_​o​p​t​m,tm​a​x>t​h​r​e​s​h​o​l​d{diff}_{not\_optm,t_{max}}>threshold and tm​a​x=tm​i​nt_{max}=t_{min} then
31:             𝕆←tm​a​x\mathbb{O}\leftarrow t_{max}.
32:          else
33:             𝕀←t\mathbb{I}\leftarrow t.
34:          end if
35:       end for
36:    end if
37:    return 𝕆,𝕀\mathbb{O},\mathbb{I}
38:end procedure