High Dynamic Range (HDR) imaging with modulo cam-eras involves solving a challenging inverse problem, where degradation occurs due to the modulo operation applied to the target HDR image.

Mod-Squad: Designing Mixtures of Experts As Modular Multi-Task Learners Zitian Chen1, Yikang Shen2, Mingyu Ding3, Zhenfang Chen2, Hengshuang Zhao3, Erik Learned-Miller1, Chuang Gan1,2

To this aim we designed a CNN structure inspired from demosaicing algorithms and directed at classifying image blocks by their position in the image modulo (2 x 2).

We evaluate our DUN, PnP-UA and UnModNet which are specifically designed for modulo HDR reconstruction, on the sample of real-sensor dataset [2]. This dataset collects 8-bit grayscale modulo …

Zhao et al. [58] used a modulo camera that is able to wrap the high radiance of dynamic range scene pe-riodically and save modulo information, then used Markov Random Field to unwrap and predict …

Robustness verification for MaxPool-based neural net-works Recently, Some work based on mixed integer lin-ear programming and satisfiability modulo theory [1, 15, 32] proposed a complete …

Positional learning trains the model to learn the modulo-2 position of pixels, leveraging the translation-invariance of CNN to replicate the underlying mosaic and its potential inconsistencies.

where “fmod” is the modulo operator. t should be integer and the “structured environment map” pattern is discrete. Specular coverage requirement. Given metallic (i.e., without diffuse part) or even mirror …

Positional learning trains the model to learn the modulo-2 position of pixels, leveraging the translation-invariance of CNN to replicate the underlying mosaic and its potential inconsistencies.

Various certification and ver-ification methods have been proposed based on Satisfia-bility Modulo theories [20, 29], mixed integer linear pro-gramming [7, 21], solving optimization problems [19, 58] …