Searcharxiv⌕ Search

arXiv subjects

Daewon Chung

Publications and source records attributed to Daewon Chung.

10 recordsLinked to original sources
arXiv2025

LiM-YOLO: Less is More with Pyramid Level Shift for Ship Detection in Optical Remote Sensing

General-purpose object detectors face fundamental structural limitations when applied to ship detection in satellite imagery, where the ship scale distribution is concentrated at small sizes and high aspect ratios. In conventional You Only Look Once architectures, the deepest feature pyramid level (stride 32) compresses narrow vessels into sub-pixel representations, causing severe spatial feature dilution and compromising accurate ship boundary regression. We propose Less is More YOLO, a streamlined detector built upon the extra-large variant of YOLOv9, to address these domain-specific structural conflicts. From a statistical analysis of ship scale distributions across four major benchmarks (SODA-A, DOTA-v1.5, FAIR1M-v2.0, and ShipRSImageNet), we introduce a Pyramid Level Shift Strategy that shifts the detection head from strides 8, 16, and 32 to strides 4, 8, and 16. This shift satisfies a spatial representability condition derived from the Nyquist-Shannon principle for the narrowest targets, while eliminating the computational redundancy of the deepest pyramid level. To further stabilize training on high-resolution satellite inputs, we incorporate a group-normalized composite-backbone projection module, mitigating gradient instability in memory-constrained micro-batch regimes. Validated on these four datasets, our detector attains an mAP50:95 of 0.600 with only 21.16 million parameters, a 64.1% reduction from the extra-large YOLOv9 baseline (58.99 million). Despite this compact size, our model surpasses state-of-the-art detectors up to three times larger, validating that a well-targeted pyramid level shift achieves a "Less is More" balance between accuracy and efficiency. The code is available at https://github.com/egshkim/LiM-YOLO.

cs.CV↗
arXiv2025

Two-weight estimates for the square function and $t$-Haar multipliers

We present necessary and sufficient conditions on triples of weights $(u,v,w)$ for the boundedness of the dyadic weighted square function $S_w$ from $L^2(u)$ into $L^2(v)$. We use this characterization to obtain necessary and sufficient conditions for the boundedness of the $t$-Haar multipliers from $L^2(u)$ into $L^2(v)$ in terms of boundedness of the dyadic weighted square function.

math.CA↗
arXiv2023

Weighted Inequalities for $t$-Haar multipliers

In this paper, we provide necessary and sufficient conditions on a triple of weights $(u,v,w)$ so that the $t$-Haar multipliers $T^t_{w,\sigma}$, $t\in \R$, %defined in \cite{P} when $\sigma=1$, are uniformly (on the choice of signs $\sigma$) bounded from $L^2(u)$ into $L^2(v)$. These dyadic operators have symbols $s(x,I)=\sigma_I\,(w(x)/\langle w\rangle_I)^t$ which are functions of the space variable $x\in\R$ and the frequency variable $I\in \mathcal{D}$, making them dyadic analogues of pseudo-differential operators. Here $\mathcal{D}$ denotes the dyadic intervals, $\sigma_I=\pm1$, and $\langle w\rangle_I$ denotes the integral average of $w$ on $I$. When $w\equiv 1$ we have the martingale transform and our conditions recover the known two-weight necessary and sufficient conditions of Nazarov, Treil and Volberg. %We will discuss some relations between the three weights inequality for these operators given the inequality for other dyadic operators. We also show how these conditions are simplified when $u=v$. In particular, the martingale one-weight and the $t$-Haar multiplier unsigned and unweighted (corresponding to $\sigma_I\equiv 1$ and $u=v\equiv 1$) known results are recovered or improved. We also obtain necessary and sufficient testing conditions of Sawyer type for the two-weight boundedness of a single variable Haar multiplier similar to those known for the martingale transform.

math.CA↗
arXiv2016

On two weight estimates for dyadic operators

We provide a quantitative two weight estimate for the dyadic paraproduct $π_b$ under certain conditions on a pair of weights $(u;v)$ and $b$ in $Carl_{u,v}$, a new class of functions that we show coincides with BMO when $u = v \in A^d_2$. We obtain quantitative two weight estimates for the dyadic square function and the martingale transforms under the assumption that the maximal function is bounded from $L_2(u)$ into $L_2(v)$ and $v \in RH^d_1$. Finally we obtain a quantitative two weight estimate from $L_2(u)$ into $L_2(v)$ for the dyadic square function under the assumption that the pair $(u; v)$ is in joint $A^d_2$ and $u^{-1} \in RH^d_1$, this is sharp in the sense that when $u = v$ the conditions reduce to $u \in A^d_2$ and the estimate is the known linear mixed estimate.

math.FA↗
arXiv2013

Cloaking due to anomalous localized resonance in plasmonic structures of confocal ellipses

If a core of dielectric material is coated by a plasmonic structure of negative dielectric material with non-zero loss parameter, then anomalous localized resonance may occur as the loss parameter tends to zero and the source outside the structure can be cloaked. It has been proved that the cloaking due to anomalous localized resonance (CALR) takes place for structures of concentric disks and the critical radius inside which the sources are cloaked has been computed. In this paper, it is proved that CALR takes place for structures of confocal ellipses and the critical elliptic radii are computed. The method of this paper uses the spectral analysis of the Neumann-Poincaré type operator associated with two interfaces (the boundaries of the core and the shell).

math.AP↗
arXiv2010

Weighted inequalities for multivariable dyadic paraproducs

Using Wilson's Haar basis in $\R^n$, which is different than the usual tensor product Haar functions, we define its associated dyadic paraproduct in $\R^n$. We can then extend "trivially" Beznosova's Bellman function proof of the linear bound in $L^2(w)$ with respect to $[w]_{A_2}$ for the 1-dimensional dyadic paraproduct. Here trivial means that each piece of the argument that had a Bellman function proof has an $n$-dimensional counterpart that holds with the same Bellman function. The lemma that allows for this painless extension we call the good Bellman function Lemma. Furthermore the argument allows to obtain dimensionless bounds in the anisotropic case.

math.FA↗
arXiv2010

Sharp bounds for general commutators on weighted Lebesgue spaces

We show that if an operator T is bounded on weighted Lebesgue space L^2(w) and obeys a linear bound with respect to the A_2 constant of the weight, then its commutator [b,T] with a function b in BMO will obey a quadratic bound with respect to the A_2 constant of the weight. We also prove that the kth-order commutator T^k_b=[b,T^{k-1}_b] will obey a bound that is a power (k+1) of the A_2 constant of the weight. Sharp extrapolation provides corresponding L^p(w) estimates. The results are sharp in terms of the growth of the operator norm with respect to the A_p constant of the weight for all 1<p<\infty, all k, and all dimensions, as examples involving the Riesz transforms, power functions and power weights show.

math.CA↗
arXiv2010

Sharp estimates for the commutators of the Hilbert, Riesz transforms and the Beurling-Ahlfors operator on weighted Lebesgue spaces

We prove that the operator norm on weighted Lebesgue space L2(w) of the commutators of the Hilbert, Riesz and Beurling transforms with a BMO function b depends quadratically on the A2-characteristic of the weight, as opposed to the linear dependence known to hold for the operators themselves. It is known that the operator norms of these commutators can be controlled by the norm of the commutator with appropriate Haar shift operators, and we prove the estimate for these commutators. For the shift operator corresponding to the Hilbert transform we use Bellman function methods, however there is now a general theorem for a class of Haar shift operators that can be used instead to deduce similar results. We invoke this general theorem to obtain the corresponding result for the Riesz transforms and the Beurling-Ahlfors operator. We can then extrapolate to Lp(w), and the results are sharp for 1 < p < 1.

math.FA↗