Studying Very Low Resolution Recognition Using Deep Networks

By Zhangyang Wang, Shiyu Chang, Yingzhen Yang, Ding Liu, Thomas S. Huang. 2016

236 citations recorded by the source corpus.

Visual recognition research often assumes a sufficient resolution of the region of interest (ROI). That is usually violated in practice, inspiring us to explore the Very Low Resolution Recognition (VLRR) problem. Typically, the ROI in a VLRR problem can be smaller than 16 16 pixels, and is challenging to be recognized even by human experts. We attempt to solve the VLRR problem using deep learning methods. Taking advantage of techniques primarily in super resolution, domain adaptation and robust regression, we formulate a dedicated deep learning method and demonstrate how these techniques are incorporated step by step. Any extra complexity, when introduced, is fully justified by both analysis and simulation results. The resulting Robust Partially Coupled Networks achieves feature enhancement and recognition simultaneously. It allows for both the flexibility to combat the LR-HR domain mismatch, and the robustness to outliers. Finally, the effectiveness of the proposed models is evaluated on three different VLRR tasks, including face identification, digit recognition and font recognition, all of which obtain very impressive performances.

Subfields: Computational Mechanics, Computer Vision and Pattern Recognition, Media Technology

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computational Mechanics

In Computer Vision and Pattern Recognition

In Media Technology

Citations in the corpus

Bowtie Networks: Generative Modeling for Joint Few-Shot Recognition and Novel-View Synthesis (2020, background): “By contrast, the performance of current well-performing recognition models substantially degrades with low-resolution images [38, 1].”
Real-World Single Image Super-Resolution: A Brief Review (2021, methodology): “higher resolution output from the LR observation, we can obtain images with the resolution beyond the limit of imaging systems, thereby benefiting the subsequent analysis and understanding tasks such as segmentation [1]–[4], detection [5]–[7], and recognition [8]–[10].”
AnomalyNet: An Anomaly Detection Network for Video Surveillance (2019, background): “tion or called differentiable programming has attracted a lot of interests [25], [42]–[48] thanks to following advantages.”
Towards Privacy-Preserving Visual Recognition via Adversarial Training: A Pilot Study (2018, methodology): “We employ state-of-the-art video recognition CNNs for corresponding tasks, and adapt them for the degraded input fd(X) using the robust pre-training strategy proposed in [61].”
AOD-Net: All-in-One Dehazing Network (2017, background): “Recently, [27, 4] validated that a joint optimization of restoration and recognition steps would boost the performance over the traditional two-stage approach.”
Low-Resolution Face Recognition in the Wild via Selective Knowledge Distillation (2018, methodology): “Following the experimental settings of [36], we choose a 180-subject subset of original-resolution images by ranking the subjects according to the number of images.”
ARDIS: a Swedish historical handwritten digit dataset (2019, methodology): “Wang et al. [41] propose a deep learning method to solve the very low-resolution digit recognition problem.”
Effects of Image Degradations to CNN-based Image Classification (2018, background): “In [13], special degradations of low image resolution was studied in the applications of face identiﬁcation, digit recognition and font recognition.”
Advancing Image Understanding in Poor Visibility Environments: A Collective Benchmark Study (2019, background): “In [40, 142], the joint optimization pipeline for low-resolution recognition”
UnitBox: An Advanced Object Detection Network (2016, background): “The ﬁrst is that in the (cid:96) 2 loss, the coordinates of a bounding box (in the form of x t , x b , x l , x r ) are optimized as four independent variables.”
Benchmarking Single-Image Dehazing and Beyond (2017, methodology): “Haze?: [17] for the first time reported the promising performance on detecting objects in the haze, by concatenating and jointly tuning AOD-Net with FRCNN as one unified pipeline, following [56], [57].”
No More Strided Convolutions or Pooling: A New CNN Building Block for Low-Resolution Images and Small Objects (2022, background): “This same requirement of high-resolution training images is also needed by several other studies such as [37].”
Single Image Deraining: A Comprehensive Benchmark Analysis (2019, background): “As pointed out by several recent works [41, 15, 42, 43], the performance of high-level computer vision tasks, such as object detection and recognition, will deteriorate in the presence of various sensory and environmental degradations.”
Low-resolution Visual Recognition via Deep Feature Distillation (2019, methodology): “Therefore, effective methods for learning new recognition models suitable for very low-resolution recognition (VLRR) problem [9] is urgently required.”
Wide Activation for Efficient and Accurate Image Super-Resolution (2018, background): “It has many applications in security, surveillance, satellite, medical imaging [24, 34] and can serve as a built-in module for other image restoration or recognition tasks [6, 21, 37, 40, 41].”