LUCES: A Dataset for Near-Field Point Light Source Photometric Stereo

Mecca, Roberto; Logothetis, Fotios; Budvytis, Ignas; Cipolla, Roberto

doi:10.48550/arxiv.2104.13135

Cited by 2 publications

(10 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As mentioned in Section 4.2 of the paper, existing photometric stereo (PS) datasets [36,49] are primarily interested in the object region, and the shadow and shading information cannot be observed in the background regions. Therefore, they are not suitable to evaluate our method in full scene reconstruction.…”

Section: E Results On the Luces Datasetmentioning

confidence: 99%

“…2). Following existing nearfield photometric stereo methods [36,46], we assume a calibrated perspective camera and known point light positions. Instead of representing the visible surface with a normal / depth map like others [25,36,46], we adopt a 3D neural field representation [3,37,41] to describe the 3D scene.…”

Section: Methodsmentioning

confidence: 99%

“…The above methods assume directional lightings. For near-field PS problem, methods based on PDE [42] and deep learning [29,32,36,46] have been proposed. More recently, Li et al [25] proposes a coordinate-based MLP to represent the normal map of the visible surface assuming directional lights.…”

Section: Related Workmentioning

confidence: 99%

“…Lighting model Following previous works [36,46], we adopt the inverse-square law for point light attenuation where light intensity L int is proportional to the multiplicative inverse of the square of the distance s (i.e., L int ∝ 1/s 2 ). The incident light direction w i and light intensity L int at a point x are given by…”

Section: Physics-based Color Renderingmentioning

confidence: 99%

“…However, existing photometric stereo datasets are mostly interested in the object region and intentionally remove the influence of the background (e.g., cover the background with black cloth to avoid interreflections [49]), which makes the shadow and shading information invisible in the background regions. Therefore, such datasets [36,49] are not suitable to evaluate the full potential of our method.…”

Section: Datasetsmentioning

confidence: 99%

See 4 more Smart Citations

S$^3$-NeRF: Neural Reflectance Field from Shading and Shadow under a Single Viewpoint

Yang¹,

Chen²,

Chen³

et al. 2022

Preprint

View full text Add to dashboard Cite

In this paper, we address the "dual problem" of multi-view scene reconstruction in which we utilize single-view images captured under different point lights to learn a neural scene representation. Different from existing single-view methods which can only recover a 2.5D scene representation (i.e., a normal / depth map for the visible surface), our method learns a neural reflectance field to represent the 3D geometry and BRDFs of a scene. Instead of relying on multi-view photoconsistency, our method exploits two information-rich monocular cues, namely shading and shadow, to infer scene geometry. Experiments on multiple challenging datasets show that our method is capable of recovering 3D geometry, including both visible and invisible parts, of a scene from single-view images. Thanks to the neural reflectance field representation, our method is robust to depth discontinuities. It supports applications like novel-view synthesis and relighting. Our code and model can be found at https://ywq.github.io/s3nerf.

show abstract

Section: E Results On the Luces Datasetmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Physics-based Color Renderingmentioning

confidence: 99%

Section: Datasetsmentioning

confidence: 99%

See 3 more Smart Citations

S$^3$-NeRF: Neural Reflectance Field from Shading and Shadow under a Single Viewpoint

Yang¹,

Chen²,

Chen³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

Industry-Fit AI Usage for Crack Detection in Ground Steel

et al. 2022

View full text Add to dashboard Cite

We investigated optimal implementation strategies for industrial inspection systems aiming to detect cracks on ground steel billets’ surfaces by combining state-of-the-art AI-based methods and classical computational imaging techniques. In 2D texture images, the interesting patterns of surface irregularities are often surrounded by visual clutter, which is to be ignored, e.g., grinding patterns. Even neural networks struggle to reliably distinguish between actual surface disruptions and irrelevant background patterns. Consequently, the image acquisition procedure already has to be optimised to the specific application. In our case, we use photometric stereo (PS) imaging to generate 3D surface models of steel billets using multiple illumination units. However, we demonstrate that the neural networks, especially in high-speed scenarios, still suffer from recognition deficiencies when using raw photometric stereo camera data, and are unable to generalise to new billets and image acquisition conditions. Only the additional application of adequate state-of-the-art image processing algorithms guarantees the best results in both aspects. The neural networks benefit when appropriate image acquisition methods together with image processing algorithms emphasise relevant surface structures and reduce overall pattern variation. Our proposed combined strategy shows a 9.25% better detection rate on validation data and is 14.7% better on test data, displaying the best generalisation.

show abstract

LUCES: A Dataset for Near-Field Point Light Source Photometric Stereo

Cited by 2 publications

References 34 publications

S$^3$-NeRF: Neural Reflectance Field from Shading and Shadow under a Single Viewpoint

S$^3$-NeRF: Neural Reflectance Field from Shading and Shadow under a Single Viewpoint

Industry-Fit AI Usage for Crack Detection in Ground Steel

Contact Info

Product

Resources

About