语料库 OCR 待核对照表

PDF 经 MinerU 转 Markdown 时丢失的符号（U+FFFD）。本表在 d29389a 之后 已逐处对照 papers/<同名>.pdf 的原文（含文本层与算法框）核定并回填。

判定原则：宁缺勿错。� 是诚实的"此处识别失败"；猜一个错的符号则是静默篡改语料。本次 245 处全部能在 PDF 上唯一确定，无一处保留 �。

两条需要读者留意的回填说明：

下标写法：原文写作 N_t / N_V / α_i / t_opt 之处，MinerU 有时把下标写成相邻的普通字符位，本表按字面字符回填（Nt / NV / αi / opt），未擅自补 _ 或 <sub>。
被读成空格的 glyph：md/2024_StopThePop… 第 148 行，原文 𝑡opt 的中间那颗 p 被读成了空格（不是 U+FFFD），已一并改回 p，此行另在备注列注明。

文件	行号	上下文	扫描建议	核定后的符号	备注
`md/2022_A_Progress_Review_on_Solid_State_LiDAR_and_Nanophotonics.md`	81	…rped sinusoidal TX and RX signal. The delay (∆�) is due to the moving of the sensing object.…		`τ`
`md/2022_A_Progress_Review_on_Solid_State_LiDAR_and_Nanophotonics.md`	112	…ivates the Pockels cell after a delay time of �. The optical beam from the laser is collimate…		`τ`
`md/2022_InstantNGP.md`	34	…>e number of parameters � and the desi…	T	`T`
`md/2022_InstantNGP.md`	60	…$ as a multiresolution sequence of � ∈ N sine and cosine functions…		`L`
`md/2022_InstantNGP.md`	119	… relative position of x within the respective �-th voxel. (4) we concatenate the resu</s…		`l`
`md/2022_InstantNGP.md`	121	…h table size � and max. resolu</s…	T	`T`
`md/2022_InstantNGP.md`	126	…Given a fully connected neural network �(y; Φ), we are interested in an encoding of it…		`m`
`md/2022_InstantNGP.md`	126	…sub>ng parameters �. These are <…	T	`θ`
`md/2022_InstantNGP.md`	126	…sub>arranged into � levels, each…	L	`L`
`md/2022_InstantNGP.md`	126	…on- taining up to � feature vectors with dimensionality</sub…	F / d	`T`
`md/2022_InstantNGP.md`	136	…er of levels �, the growth fac</s…	L	`L`
`md/2022_InstantNGP.md`	140	…⌊x�⌋ and ⌈x�⌉ span a …		`l`
`md/2022_InstantNGP.md`	140	…⌊x�⌋ and ⌈x�⌉ span a voxel with $2 ^ { d …		`l`
`md/2022_InstantNGP.md`	140	…ize ofat most�. For coarse leve</…		`T`
`md/2022_InstantNGP.md`	140	…nse grid requires fewer than � parameters, i.e. $( N _…	T	`T`
`md/2022_InstantNGP.md`	148	…>y $d - 1$ of the � dimensions must be…	F / d	`d`
`md/2022_InstantNGP.md`	150	…b>each corner are �-linearly i<s…		`d`
`md/2022_InstantNGP.md`	152	…ub>or each of the � levels. The …	L	`L`
`md/2022_InstantNGP.md`	154	…er values of � result in higher qualit…		`T`
`md/2022_InstantNGP.md`	154	… We recommend practitioners to use � to tweak the enco<…		`T`
`md/2022_InstantNGP.md`	161	…h table size � which det<su…	T	`T`
`md/2022_InstantNGP.md`	170	…The hyperparameters � (number of levels) and � (number of feature d…	T	`L`
`md/2022_InstantNGP.md`	170	…sub>parameters � (number of levels) and � (number of feature dimensions) also trade of …	T	`F`
`md/2022_InstantNGP.md`	170	…coding parameters � in Figure 5. In this analy</s…	T	`θ`
`md/2022_InstantNGP.md`	172	…sub>ual integer coordinates ⌊x�⌋ are not considered a collision; a coll…		`l`
`md/2022_InstantNGP.md`	180	…s ensures that the encoding enc(x; �), and by the chain rule its compos…		`θ`
`md/2022_InstantNGP.md`	180	… rule its composition with the neural network �(enc(x; θ); Φ), are continuous. Without inter<…		`m`
`md/2022_InstantNGP.md`	180	…ctions, in which case the gradient ��(enc(x; θ); Φ)/�x, i.e. the surface normal, <s…		`∂m`	相邻两个字符位：偏导算子 ∂ 与网络函数名 m
`md/2022_InstantNGP.md`	180	…> in which case the gradient ��(enc(x; θ); Φ)/�x, i.e. the surface normal, would i…		`∂`
`md/2022_InstantNGP.md`	192	…ub>ure dimensions � per lookup d…	F / d	`F`
`md/2022_InstantNGP.md`	192	…and, a large � favors memory coherence…		`F`
`md/2022_InstantNGP.md`	225	…b>with a fixed growth factor � = 2. To allow meaningful comp…		`b`
`md/2022_InstantNGP.md`	239	…ing application [Müller et al. 2021]. The MLP � enc(�; �); Φ is tasked with predicting phot…		`m`
`md/2022_InstantNGP.md`	239	…lication [Müller et al. 2021]. The MLP � enc(�; �); Φ is tasked with predicting photorealis…		`x`
`md/2022_InstantNGP.md`	239	…ation [Müller et al. 2021]. The MLP � enc(�; �); Φ is tasked with predicting photorealistic…		`θ`
`md/2022_InstantNGP.md`	241	…Multiresolution hash encodin (Ours), � = 15, 133 FPS …		`T`
`md/2022_InstantNGP.md`	251	…onal features as auxiliary encoded dimensions � to be concatenated with the encoded position,…	F / d	`ξ`
`md/2022_InstantNGP.md`	308	…her than reduce (for example, by summing) the � -di-mensional fea<…		`F`
`md/2022_InstantNGP.md`	308	…> dimensionality of the encoded result y from �� to � may be too small to enco…	F / d	`LF`	相邻两个字符位：L 与 F（原文写作 𝐿𝐹）
`md/2022_InstantNGP.md`	308	…nsionality of the encoded result y from �� to � may be too small to enco</sub…		`F`
`md/2022_InstantNGP.md`	308	…b>ormation. While � could be increase<…		`F`
`md/2022_InstantNGP.md`	310	…utational cost of increasing � could be insignificant. We thus ar…		`F`
`md/2022_InstantNGP.md`	310	…tions, concatenation, coupled with � = 2 always yielded b<su…		`F`
`md/2022_InstantNGP.md`	316	…he least significant bits of Z� by a space-fillin…		`d`
`md/2022_InstantNGP.md`	316	…lly better than setting �1 := 1 as done in</…		`π`
`md/2022_InstantNGP.md`	452	…re smoother interpolation than the �-linear interpolation th…		`d`
`md/2022_InstantNGP.md`	454	…sub> using a �-quadratic <s…		`d`
`md/2022_InstantNGP.md`	454	…ub>dratic or �-cubic int<su…		`d`
`md/2022_InstantNGP.md`	460	…to the �-linear int<s…		`d`
`md/2022_InstantNGP.md`	500	…o compute) with standard deviation �/1024, where � is the bounding <sub…		`r`
`md/2022_InstantNGP.md`	500	…ub>ute) with standard deviation �/1024, where � is the bounding radi<su…		`r`
`md/2022_InstantNGP.md`	540	… step size proportional to the distance � along the ray $Δ t : = t / 256$ c…		`t`
`md/2022_InstantNGP.md`	540	…ar { 2 } 4 } \right]$ , where � is size of th<sub…		`s`
`md/2022_InstantNGP.md`	540	…s exponential growth in �, which means that the comp</s…		`t`
`md/2022_InstantNGP.md`	548	…}$ with a constant number of levels (variable � as in Equation (3)) or log</s…	L	`L`
`md/2022_InstantNGP.md`	548	…rithmically many levels � (constant �).…	L	`L`
`md/2022_InstantNGP.md`	548	…ub>y many levels � (constant �).…	L	`b`
`md/2022_InstantNGP.md`	552	…/sub>ace, we maintain a cascade of � multiscale occupancy gr…		`K`
`md/2022_InstantNGP.md`	552	… occupancy grids, where � = 1 for all synthetic NeRF scenes …		`K`
`md/2022_InstantNGP.md`	556	…sub>ch one of the � grids is quer</sub…		`K`
`md/2022_InstantNGP.md`	556	…>e sample position x and the step size Δ�: among the grids covering x, the finest one w…		`t`
`md/2022_InstantNGP.md`	556	…ide-length larger than Δ� is queried.…		`t`
`md/2022_InstantNGP.md`	564	…(2) randomly sample � candidate cells, and set their value to the <…		`M`
`md/2022_InstantNGP.md`	568	…>tegy of the � candidate <s…		`M`
`md/2023_3DGS.md`	108	…ats allowing fast �-blending for…	α	`α`
`md/2023_3DGS.md`	110	…01a] centered at point (mean) �:…		`μ`
`md/2023_3DGS.md`	118	…ub>formation � the covarian…		`W`
`md/2023_3DGS.md`	142	…>o positions �, �, and cova…		`p`
`md/2023_3DGS.md`	142	…positions �, �, and covaria…		`α`
`md/2023_3DGS.md`	142	…ents representing color � of each Gaussian to correctly ca<s…	C / c	`c`
`md/2023_3DGS.md`	156	…ub>ction for � to cons</sub…		`α`
`md/2023_3DGS.md`	180	…sub> is to set the � value close t<sub…		`α`
`md/2023_3DGS.md`	180	…ub>n increases the � for the Gaus…		`α`
`md/2023_3DGS.md`	180	…ub> Gaussians with � less than $\epsil…		`α`
`md/2023_3DGS.md`	187	…b>llow approximate �-blending – including</s…	α	`α`
`md/2023_3DGS.md`	189	…pixel that hindered previous �-blending solutions [Kop…	α	`α`
`md/2023_3DGS.md`	197	…each point’s � in our back-…		`α`
`md/2023_3DGS.md`	454	…ation matrix �:…		`R`
`md/2023_3DGS.md`	460	…omponents of �:…		`q`
`md/2023_3DGS.md`	473	…�, ℎ: width an…		`w`
`md/2023_3DGS.md`	474	…� ← SfM Points ⊲ Positi<s…		`M`
`md/2023_3DGS.md`	475	…�, �, � ← InitAttributes() ⊲ Co…		`S`
`md/2023_3DGS.md`	475	…�, �, � ← InitAttributes() ⊲ Cova<…		`C`
`md/2023_3DGS.md`	475	…�, �, � ← InitAttributes() ⊲ Cova</su…		`A`
`md/2023_3DGS.md`	476	…� ← 0 ⊲ Iterat…		`i`
`md/2023_3DGS.md`	478	…�, ˆ� ← SampleTrainin<sub…		`V`
`md/2023_3DGS.md`	478	…�, ˆ� ← SampleTrainingView() …		`I`
`md/2023_3DGS.md`	478	…TrainingView() ⊲ Camera � and Image …		`V`
`md/2023_3DGS.md`	479	…� ← Rasterize(�, �, �, �, �) ⊲ Al<su…		`I`
`md/2023_3DGS.md`	479	…� ← Rasterize(�, �, �, �, �) ⊲ Alg. 2 …		`M`
`md/2023_3DGS.md`	479	…� ← Rasterize(�, �, �, �, �) ⊲ Alg. 2 …		`S`
`md/2023_3DGS.md`	479	…� ← Rasterize(�, �, �, �, �) ⊲ Alg. 2 …		`C`
`md/2023_3DGS.md`	479	…� ← Rasterize(�, �, �, �, �) ⊲ Alg. 2 …		`A`
`md/2023_3DGS.md`	479	…� ← Rasterize(�, �, �, �, �) ⊲ Alg. 2 …		`V`
`md/2023_3DGS.md`	480	…� ← �� (�, ˆ�) ⊲ L…		`L`
`md/2023_3DGS.md`	480	…� ← �� (�, ˆ�) ⊲ Loss</su…		`Loss`	相邻四个字符位，组成函数名 Loss
`md/2023_3DGS.md`	480	…� ← �� (�, ˆ�) ⊲ Loss …		`I`
`md/2023_3DGS.md`	480	…� ← �� (�, ˆ�) ⊲ Loss …		`I`
`md/2023_3DGS.md`	481	…�, �, �, � ← Adam(∇�) ⊲ Backpro…		`M`
`md/2023_3DGS.md`	481	…�, �, �, � ← Adam(∇�) ⊲ Backprop</…		`S`
`md/2023_3DGS.md`	481	…�, �, �, � ← Adam(∇�) ⊲ Backprop</sub…		`C`
`md/2023_3DGS.md`	481	…�, �, �, � ← Adam(∇�) ⊲ Backprop &…		`A`
`md/2023_3DGS.md`	481	…�, �, �, � ← Adam(∇�) ⊲ Backprop & Step…		`L`
`md/2023_3DGS.md`	482	…if IsRefinementIteration(�) then …		`i`
`md/2023_3DGS.md`	483	…for all Gaussians (�, Σ, �, �) in (�, �, �, �) do …		`μ`
`md/2023_3DGS.md`	483	…for all Gaussians (�, Σ, �, �) in (�, �, �, �) do …		`c`
`md/2023_3DGS.md`	483	…for all Gaussians (�, Σ, �, �) in (�, �, �, �) do …		`α`
`md/2023_3DGS.md`	483	…for all Gaussians (�, Σ, �, �) in (�, �, �, �) do …		`M`
`md/2023_3DGS.md`	483	…for all Gaussians (�, Σ, �, �) in (�, �, �, �) do …		`S`
`md/2023_3DGS.md`	483	…for all Gaussians (�, Σ, �, �) in (�, �, �, �) do …		`C`
`md/2023_3DGS.md`	483	…for all Gaussians (�, Σ, �, �) in (�, �, �, �) do …		`A`
`md/2023_3DGS.md`	484	…if � ⊲ Pr<…		`α`
`md/2023_3DGS.md`	484	…if � ⊲ Pr…		`ε`
`md/2023_3DGS.md`	484	…if � ⊲ Pruning…		`μ`
`md/2023_3DGS.md`	488	…if ∥� ∥ $& g t; τ_{S}$ then ⊲ Ove…		`S`
`md/2023_3DGS.md`	489	…SplitGaussian(�, Σ, �, �) …	g / i / N	`μ`
`md/2023_3DGS.md`	489	…SplitGaussian(�, Σ, �, �) …	g / i / N	`c`
`md/2023_3DGS.md`	489	…SplitGaussian(�, Σ, �, �) …	g / i / N	`α`
`md/2023_3DGS.md`	491	…CloneGaussian(�, Σ, �, �) …	g / i / N	`μ`
`md/2023_3DGS.md`	491	…CloneGaussian(�, Σ, �, �) …	g / i / N	`c`
`md/2023_3DGS.md`	491	…CloneGaussian(�, Σ, �, �) …	g / i / N	`α`
`md/2023_3DGS.md`	496	…� ← � + 1 …		`i`
`md/2023_3DGS.md`	496	…� ← � + 1 …		`i`
`md/2023_3DGS.md`	505	…�, ℎ: width and h<s…		`w`
`md/2023_3DGS.md`	506	…�, �: Gaussian…	g / i / N	`M`
`md/2023_3DGS.md`	506	…�, �: Gaussian mea</sub…	g / i / N	`S`
`md/2023_3DGS.md`	507	…�, �: Gaussian…	g / i / N	`C`
`md/2023_3DGS.md`	507	…�, �: Gaussian co…	g / i / N	`A`
`md/2023_3DGS.md`	508	…�: view configuration of current cam…	v / V	`V`
`md/2023_3DGS.md`	509	…function Rasterize(�, ℎ, �, �, �, �, �)…		`w`
`md/2023_3DGS.md`	509	…function Rasterize(�, ℎ, �, �, �, �, �)…		`M`
`md/2023_3DGS.md`	509	…function Rasterize(�, ℎ, �, �, �, �, �)…		`S`
`md/2023_3DGS.md`	509	…function Rasterize(�, ℎ, �, �, �, �, �)…		`C`
`md/2023_3DGS.md`	509	…function Rasterize(�, ℎ, �, �, �, �, �)…		`A`
`md/2023_3DGS.md`	509	…function Rasterize(�, ℎ, �, �, �, �, �)…		`V`
`md/2023_3DGS.md`	510	…CullGaussian(�, �) ⊲ Frustu…	g / i / N	`p`
`md/2023_3DGS.md`	510	…CullGaussian(�, �) ⊲ Frustum C…	g / i / N	`V`
`md/2023_3DGS.md`	511	…�′, �′ ← Screensp</s…		`M`
`md/2023_3DGS.md`	511	…�′, �′ ← ScreenspaceGaussians…		`S`
`md/2023_3DGS.md`	511	…sup>′ ← ScreenspaceGaussians(�, �, � ) ⊲ Transfor…		`M`
`md/2023_3DGS.md`	511	…>′ ← ScreenspaceGaussians(�, �, � ) ⊲ Transform</…		`S`
`md/2023_3DGS.md`	511	…/sup> ← ScreenspaceGaussians(�, �, � ) ⊲ Transform</sub…		`V`
`md/2023_3DGS.md`	512	…� ← CreateTiles(�, ℎ)…		`T`
`md/2023_3DGS.md`	512	…� ← CreateTiles(�, ℎ)…		`w`
`md/2023_3DGS.md`	513	…�, � ← DuplicateWithKeys(…		`L`
`md/2023_3DGS.md`	513	…�, � ← DuplicateWithKeys(�<s…		`K`
`md/2023_3DGS.md`	513	…, � ← DuplicateWithKeys(�′, �) ⊲ Indices and Ke<s…		`M`
`md/2023_3DGS.md`	513	…/sub>licateWithKeys(�′, �) ⊲ Indices and Keys…		`T`
`md/2023_3DGS.md`	514	…SortByKeys(�, �) ⊲ Global…		`K`
`md/2023_3DGS.md`	514	…SortByKeys(�, �) ⊲ Globall<s…		`L`
`md/2023_3DGS.md`	515	…� ← IdentifyTileRanges(� …		`R`
`md/2023_3DGS.md`	515	…� ← IdentifyTileRanges(� , �)…		`T`
`md/2023_3DGS.md`	515	… IdentifyTileRanges(� , �)…		`K`
`md/2023_3DGS.md`	516	…� ← 0 ⊲ Init Canvas<…		`I`
`md/2023_3DGS.md`	517	…for all Tiles � in � do…		`t`
`md/2023_3DGS.md`	517	…for all Tiles � in � do…		`I`
`md/2023_3DGS.md`	518	…for all Pixels � in � do…		`i`
`md/2023_3DGS.md`	518	…for all Pixels � in � do…		`t`
`md/2023_3DGS.md`	519	…� ← GetTileRange(�, �)…		`r`
`md/2023_3DGS.md`	519	…� ← GetTileRange(�, �)…		`R`
`md/2023_3DGS.md`	519	…� ← GetTileRange(�, �)…		`t`
`md/2023_3DGS.md`	520	…�[�] ← BlendInOrder(�, �, �, �, �′, …	α	`I`
`md/2023_3DGS.md`	520	…�[�] ← BlendInOrder(�, �, �, �, �′, �<…	α	`i`
`md/2023_3DGS.md`	520	…�[�] ← BlendInOrder(�, �, �, �, �′, �′, �, �)…	α	`i`
`md/2023_3DGS.md`	520	…�[�] ← BlendInOrder(�, �, �, �, �′, �′, �, �)…	α	`L`
`md/2023_3DGS.md`	520	…�[�] ← BlendInOrder(�, �, �, �, �′, �′, �, �)…	α	`r`
`md/2023_3DGS.md`	520	…�[�] ← BlendInOrder(�, �, �, �, �′, �′, �, �)…	α	`K`
`md/2023_3DGS.md`	520	…�[�] ← BlendInOrder(�, �, �, �, �′, �′, �, �)…	α	`M`
`md/2023_3DGS.md`	520	…[�] ← BlendInOrder(�, �, �, �, �′, �′, �, �)…	α	`S`
`md/2023_3DGS.md`	520	…der(�, �, �, �, �′, �′, �, �)…		`C`
`md/2023_3DGS.md`	520	…(�, �, �, �, �′, �′, �, �)…		`A`
`md/2023_3DGS.md`	523	…return �…		`I`
`md/2023_3DGS.md`	527	…pass by each Gaussian’s �. Implemented naïvely,</…	g / i / N	`α`
`md/2023_3DGS.md`	527	…b>ng updates with � < � (we choose � as $\textstyle { \frac { 1 }…		`α`
`md/2023_3DGS.md`	527	… updates with � < � (we choose � as $\textstyle { \frac { 1 } { 2…		`ε`
`md/2023_3DGS.md`	527	…b>ates with � < � (we choose � as $\frac{1}{255})$ <…		`ε`
`md/2023_3DGS.md`	527	…d also clamp � with 0.99 from ab…		`α`
`md/2024_Reducing_the_Memory_Footprint_of_3D_Gaussian_Splatting.md`	84	…t. For a given view v, the pixel footprint at � identifies the world-space extent of a pixel …	v / V	`g`
`md/2024_Reducing_the_Memory_Footprint_of_3D_Gaussian_Splatting.md`	89	…ixel footprint $a_{m i n}$ at �, which yields a spatial extent proportional t…		`g`
`md/2024_Reducing_the_Memory_Footprint_of_3D_Gaussian_Splatting.md`	137	…e required memory for these attributes from 56� × 4 bytes to $5 6 N + 2 0 \times 2 5 6 \times…		`N`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	74	…�(r) is the output color for a given ray r, � (…	C / c	`c`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	74	…�(r) is the output color for a given ray r, � (r, �) is the opacity along the ray and � (r,…	α	`σ`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	74	… is the output color for a given ray r, � (r, �) is the opacity along the ray and � (r, �) is…	α	`t`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	74	… r, � (r, �) is the opacity along the ray and � (r, �) is the view-dependent emitted radiance…	α	`c`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	74	…(r, �) is the opacity along the ray and � (r, �) is the view-dependent emitted radiance. 3DGS…	α	`t`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	80	…� is the Gaussian’s location, � is a rotation m…	g / i / N	`μ`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	80	…� is the Gaussian’s location, � is a rotation matrix and � is a diagonal scal…	g / i / N	`R`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	80	…ussian’s location, � is a rotation matrix and � is a diagonal scaling matrix, allowing to pos…		`S`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	98	…where � iterates over the $N_{r}$ Gau…		`i`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	98	…ce the ray in the ordering of $t_{i},$ and �� is the opacity of the Gaussian along the ray,…	α	`αi`	相邻两个字符位：α 与下标 i（原文 𝛼𝑖）
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	100	… } $i s i n d e p e n d e n t o f r, a g l o b a l s o r t o f a l l � i s p o s s i b l e . N a ï v e l y, t h i s w o u l d l e a d t o$ N _…		`t`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	127	…iewspace z-coordinate of each Gaussian’s mean �, see Eqn. (2). This leads to a consistent sor…	g / i / N	`μ`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	148	…ditionally, see Fig. 3. Therefore, we compute �� on a per-ray basis.…		`t`	原文为 𝑡opt；其中 `p` 那一颗 glyph 被 MinerU 读成了空格（不是 U+FFFD），本次一并改回 `p`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	148	…ally, see Fig. 3. Therefore, we compute �� on a per-ray basis.…		`o`	原文为 𝑡opt；其中 `p` 那一颗 glyph 被 MinerU 读成了空格（不是 U+FFFD），本次一并改回 `p`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	148	…ly, see Fig. 3. Therefore, we compute �� on a per-ray basis.…		`t`	原文为 𝑡opt；其中 `p` 那一颗 glyph 被 MinerU 读成了空格（不是 U+FFFD），本次一并改回 `p`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	167	…ut of order, we accumulate their diference in ��. We present a visual example in…		`t`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	167	…order, we accumulate their diference in ��. We present a visual example in Fig. 4,…		`opt`	原文 𝑡opt，下标 opt 按字面回填
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	167	…le in Fig. 4, with corresponding runtimes and � in Tab. 1 — evidently, even though � decrease…		`δ`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	167	…imes and � in Tab. 1 — evidently, even though � decreases with a larger resorting window, the…		`δ`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	194	…ow \hat { \mathbf { x } } = \mu _ { 2 }$ . If �2 ∉ �, then xˆ must lie on one of t…		`μ`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	194	…f { x } } = \mu _ { 2 }$ . If �2 ∉ �, then xˆ must lie on one of the two tile edge…		`X`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	198	…X } G _ { 2 } ( { \bf x } ) t _ { o p t }$ (x)�x is a better estimate. Yet, even a numerical …		`d`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	210	…the 2×2 tile. We evaluate $t_{o p t}$ and � for the respective rays and insert the newly …		`α`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	228	…is the accumulated output color including the �-th Gaussian in front-to-back order. As we per…	C / c	`i`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	245	… frame $F_{i + t}$ with ofset � using optical flow predictions from state-of-…		`t`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	268	…on $(ϕ (μ; o, d) =$ �� , cf. Eqn. (4)).…		`t`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	268	…\phi ( \mu ; { \bf o } , { \bf d } ) =$ �� , cf. Eqn. (4)).…		`opt`	原文 𝑡opt，下标 opt 按字面回填
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	296	… despite their comparable number of Gaussians �. Even though the Train scene contains slightl…		`N`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	310	…isible (in view-frustum) number of Gaussians (� vs. ��), as well as standard deviation and av…	v / V	`N`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	310	… (in view-frustum) number of Gaussians (� vs. ��), as well as standard deviation and average n…	v / V	`NV`	相邻两个字符位：N 与下标 V（原文 𝑁𝑉）
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	310	…16×16 tiles covered by each visible Gaussian (��). We additionally include an appr…	g / i / N	`N`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	310	…tiles covered by each visible Gaussian (��). We additionally include an approximat…	g / i / N	`t`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	310	…lude an approximate number of sort entries as �� · avg(�� ).…		`NV`	相邻两个字符位：N 与下标 V（原文 𝑁𝑉）
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	310	…pproximate number of sort entries as �� · avg(�� ).…		`N`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	310	…mate number of sort entries as �� · avg(�� ).…		`t`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	313	…culling results in a larger reduction of avg. �� for Train than Bonsai (∼3.5× vs. …		`N`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	313	…g results in a larger reduction of avg. �� for Train than Bonsai (∼3.5× vs. ∼2.5×)…		`t`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	455	… on one of these two edges. The fact that the � coordinate of $d_{x}$ and the…		`y`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	455	…coordinate of $d_{x}$ and the � coordinate of $d_{y}$ are zero, a…		`x`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	459	…�min, �max, �min<su…		`x`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	459	…�min, �max, �min, $y _ { …		`x`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	459	…�min, �max, �min, $y _ { \mathrm { { m a x } } }…		`y`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	471	…if $μ_{2} Φ_{y}$ �max the…		`y`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	471	…\Phi } _ { y } $ �max then …		`y`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	487	…ma } } ^ { - 1 } $r e l y i n g o n t h e s p e c i f i c s o f � a n d �, a n d p r e c o m p u t e$ \Sigma ^ { - 1 } ( \bo…		`R`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	487	… ^ { - 1 } $r e l y i n g o n t h e s p e c i f i c s o f � a n d �, a n d p r e c o m p u t e$ \Sigma ^ { - 1 } ( \boldsymb…		`S`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	503	…y error grows almost linearly with increasing �. Further, our method outperforms 3DGS for all…		`t`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	505	…Table 8. FLIPt comparison for � ∈ {1, 3, 5, 7, 9} for three camera paths for …		`t`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	505	…be seen, our method outperforms 3DGS for each �, and FLIPt scales almost linearly …		`t`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	505	… scales almost linearly with increasing �.…		`t`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	523	…as equal. We mapped those answers onto scores �:…		`s`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	560	…nd, therefore, has a significantly lower avg. �� for our method (2.7 vs. 2.41) com…		`N`
`md/2024_StopThePop__Sorted_Gaussian_Splatting_for_View-Consisten.md`	560	…erefore, has a significantly lower avg. �� for our method (2.7 vs. 2.41) compared …		`t`

合计 245 处，全部已核定并回填。

本页由 outputs/语料OCR待核对照表.md 预渲染生成。公式以 MathML 呈现，浏览器原生渲染，不依赖外部脚本或字体 CDN。