English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Occlusion issue is the biggest challenge in pedestrian detection. RCNN-baseddetectors extract instance features by cropping rectangle regions of interestin the feature maps. However, the visible pixels of the occluded objects arelimited, making the rectangle instance feature mixed with a lot ofinstance-irrelevant noise information. Besides, by counting the number ofinstances with different degrees of overlap of CrowdHuman dataset, we find thatthe number of severely overlapping objects and the number of slightlyoverlapping objects are unbalanced, which may exacerbate the challenges posedby occlusion issues. Regarding to the noise issue, from the perspective ofdenoising, an iterable dynamic instance noise filter (DINF) is proposed for theRCNN-based pedestrian detectors to improve the signal-noise ratio of theinstance feature. Simulating the wavelet denoising process, we use the instancefeature vector to generate dynamic convolutional kernels to transform the RoIsfeatures to a domain in which the near-zero values represent the noiseinformation. Then, soft thresholding with channel-wise adaptive thresholds isapplied to convert the near-zero values to zero to filter out noiseinformation. For the imbalance issue, we propose an IoU-Focal factor (IFF) tomodulate the contributions of the well-regressed boxes and the bad-regressedboxes to the loss in the training process, paying more attention to theminority severely overlapping objects. Extensive experiments conducted onCrowdHuman and CityPersons demonstrate that our methods can help RCNN-basedpedestrian detectors achieve state-of-the-art performance.

DINF: Dynamic Instance Noise Filter for Occluded Pedestrian Detection