# PixelNet: Towards a General Pixel-level Architecture

arXiv: https://arxiv.org/abs/1609.06694 · HTML: https://mlsanity.com/paper/1609.06694/

## Key facts

- Cited by 9 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 21 Sep 2016.

## Built on by (most cited first, 9 of 9)

- [Shape from Polarization for Complex Scenes in the Wild](https://mlsanity.com/paper/2112.11377/) (2021)
- [Divide-and-Conquer for Lane-Aware Diverse Trajectory Prediction](https://mlsanity.com/paper/2104.08277/) (2021)
- [Semantic Segmentation with Labeling Uncertainty and Class Imbalance](https://mlsanity.com/paper/2102.04566/) (2021)
- [CARLA2Real: a tool for reducing the sim2real appearance gap in CARLA simulator](https://mlsanity.com/paper/2410.18238/) (2024)
- [ORDNet: Capturing Omni-Range Dependencies for Scene Parsing](https://mlsanity.com/paper/2101.03929/) (2021)
- [MT: Multi-Perspective Feature Learning Network for Scene Text Detection](https://mlsanity.com/paper/2105.05455/) (2021)
- [Artificial intelligence based prediction on lung cancer risk factors using deep learning](https://mlsanity.com/paper/2304.05065/) (2023)
- [Bird Eye-View to Street-View: A Survey](https://mlsanity.com/paper/2405.08961/) (2024)
- [Diffusion-Driven Two-Stage Active Learning for Low-Budget Semantic Segmentation](https://mlsanity.com/paper/2510.22229/) (2025)

## Abstract (arXiv)

We explore architectures for general pixel-level prediction problems, from low-level edge detection to mid-level surface normal estimation to high-level semantic segmentation. Convolutional predictors, such as the fully-convolutional network (FCN), have achieved remarkable success by exploiting the spatial redundancy of neighboring pixels through convolutional processing. Though computationally efficient, we point out that such approaches are not statistically efficient during learning precisely because spatial redundancy limits the information learned from neighboring pixels. We demonstrate that (1) stratified sampling allows us to add diversity during batch updates and (2) sampled multi-scale features allow us to explore more nonlinear predictors (multiple fully-connected layers followed by ReLU) that improve overall accuracy. Finally, our objective is to show how a architecture can get performance better than (or comparable to) the architectures designed for a particular task. Interestingly, our single architecture produces state-of-the-art results for semantic segmentation on PASCAL-Context, surface normal estimation on NYUDv2 dataset, and edge detection on BSDS without contextual post-processing.