From 2b7b21416c7b97a7188503f99a01eb05c26ec440 Mon Sep 17 00:00:00 2001 From: Dustella Date: Mon, 21 Apr 2025 10:44:47 +0800 Subject: [PATCH] fix: correct some info --- assignment_2/slide.typ | 47 +++++++++++++++++++----------------------- 1 file changed, 21 insertions(+), 26 deletions(-) diff --git a/assignment_2/slide.typ b/assignment_2/slide.typ index 9098379..787607a 100644 --- a/assignment_2/slide.typ +++ b/assignment_2/slide.typ @@ -61,7 +61,7 @@ Given a point cloud *$P = \{p_1, p_2, dots, p_n\}$* where each point *$p_i in RR^3$* represents a 3D coordinate in the railway environment, -our task is to assign each point a semantic label *$l_i in \{0, 1, dots, C-1\}$* +Our task is to assign each point a semantic label *$l_i in \{0, 1, dots, C-1\}$* where *$C = 13$* represents our predefined classes. @@ -77,7 +77,7 @@ The function $f: P -> L$ maps the input point cloud to a set of labels $L = \{l_ #v(1em) #text(size: 1.2em, weight: "bold")[Pilot Study Aims] - - Establish baseline performance using RandLA-Net + - Establish *baseline* performance using RandLA-Net - Evaluate feasibility of detecting extremely rare objects (0.001% of data) ][ @@ -109,32 +109,26 @@ The function $f: P -> L$ maps the input point cloud to a set of labels $L = \{l_ #text(size: 1.1em, weight: "bold")[Training Setup] - - 858 training files, 172 test files - - Only 18 training files and 1 test file contain foreign objects - - 1/4 downsampling ratio + - *858* training files, *172* test files + - Only *18* training files and *1* test file contain foreign objects + - *1/4* downsampling ratio - NVIDIA RTX 4090 GPU ][ #text(size: 1.1em, weight: "bold")[Data Collection] - - 1,031 PLY files with 248M+ points - - 13 semantic classes including railway infrastructure elements - - "Box" class (label 11) represents foreign objects - - Extreme class imbalance: boxes only 0.001% of points + - *1,031* PLY files with *248M+* points + - *13* semantic classes including railway infrastructure elements + - "Box" class (label *11*) represents foreign objects + - Extreme class imbalance: boxes only *0.001%* of points ] -#slide(composer: (1fr, 3fr))[ - #text(size: 1em, weight: "bold")[Inspection on data] +#slide(composer: (2fr, 3fr))[ + #text(size: 1.2em, weight: "bold")[Inspection on data] #v(1em) #text(size: 0.8em)[Table: Distribution of semantic classes in the railway LiDAR dataset] ][ - - - - // #text(size: 1.1em, weight: "bold")[Inspection on Data] - - #set text(size: 0.7em) #table( columns: (auto, auto, auto, auto), @@ -153,7 +147,7 @@ The function $f: P -> L$ maps the input point cloud to a set of labels $L = \{l_ [8], [Mountain], [51,685,366], [20.82%], [9], [Train], [9,047,963], [3.65%], [10], [Human], [275,077], [0.11%], - [11], [Box (foreign object)], [3,080], [0.001%], + [11], [*Box (object)*], [*3,080*], [*0.001%*], [12], [Others], [2,360,810], [0.95%], ) @@ -190,7 +184,7 @@ The function $f: P -> L$ maps the input point cloud to a set of labels $L = \{l_ #slide()[ #text(weight: "bold")[Results] - - The overall mean IoU across all classes was 70.29\%, + - The overall mean IoU across all classes was *70.29\%*, - the IoU for our target class— *"Box"* (foreign object)—was *0.00\%* (Will be discussed later). - IoU of other classes was relatively high, with *"Train"* achieving *95.22\%* and *"Ground"* achieving *89.68\%*. ][ @@ -237,15 +231,14 @@ The function $f: P -> L$ maps the input point cloud to a set of labels $L = \{l_ Let's look at Cross Entropy Loss: - // \ell(x, y) = L = \{l_1,\dots,l_N\}^\top, \quad - // l_n = - w_{y_n} \log \frac{\exp(x_{n,y_n})}{\sum_{c=1}^C \exp(x_{n,c})} - // \cdot \mathbb{1}\{y_n \not= \text{ignore\_index}\} $ ell(x, y) = 1 / N sum_(n=1)^N - w_(y_n) log frac(exp(x_{n,y_n}) , sum_(c=1)^C exp(x_{n,c})) $ where $w_{y_n}$ is the weight for class $y_n$ and $N$ is the number of points in the batch. + In this case we just *blindly set weight for all class as 1*, which is not suitable. We should add weight on classes like boxes and human. + ] @@ -263,7 +256,9 @@ The function $f: P -> L$ maps the input point cloud to a set of labels $L = \{l_ #slide(align: auto)[ - #v(6em) + #text(weight: "bold")[Why Model Performs Bad on Boxes] + + #v(5em) The model is *biased towards the majority classes*, leading to poor performance on the minority class (foreign objects). @@ -273,9 +268,9 @@ The function $f: P -> L$ maps the input point cloud to a set of labels $L = \{l_ #text(weight: "bold")[Future Work] - - Add weights in loss functions to address class imbalance - - Explore data augmentation techniques to increase the representation of the "Box" class - - Consider ensemble methods or multi-task learning to improve detection performance + - Add weights in loss functions to *address class imbalance* + - Explore *data augmentation* techniques to increase the representation of the "Box" class + - Consider *ensemble methods* or multi-task learning to improve detection performance ]