Impact research suggested for international publication

Nov 6, 2024, International Office，5-minute read  

Recently, the Nature Research paper: Apply prior feature integration to sparse object detectors by QIAN Yu, a graduate student who is studying for AXHU-based Joint Master’s program, which was supervised by AXHU Prof. WANG Qijin (also as corresponding author), was officially publishedin Pattern Recognition (SCI).

For more research, go to https://doi.org/10.1016/j.patcog.2024.111103

The evidence-based journal [Impact factor 7.5/CiteScore 14.4/ISSN 0031-3203] created by Elsevier Inc. of Bristish RELX Group was also included into the Top CS Journal catalog of the Chinese Academy of Sciences, which theoretically aligns the data-centric access and transformation with international dynamics and consistency-driven feature in the cognate scope of computer vision and Artificial Intelligence for supporting accountability of metrics and rankings in 2024.

When assessing academic career progress and other indicators, which has become the one of the Class-B Journal Recommendations by the China Computer Federation (CCF), and of the Class-A at the Chinese Association of Automation(CAA).

Grated at efforts by “noise frame” with the aid of Gaussian-derived processes available for sparse object detection to address some critical questions of noisy boxes such as global features capturing and low-efficient matching in the Feature Pyramid Network (FPN), this research aims to solve above the two challenges via the Prior Sparse R-CNN embedded with Aggregated Encoder required for signal detection and graph-based techniques not only facing the object scale-up through the dilated residual blocks and feature aggregation, but underpinning its workout in the condition of one feature map.  

Specifically, Prior Sparse R-CNN had introduced the Region Generation Network(RGN), which can well help generate the prior prediction generation via the extra data-based training in the fields.

Combining the edition of information systems, empirically precisely, with the noisy boxes sampling, this design, being in contrast to the existing solutions and approaches, swoops in a capacity of trending performance and high efficiency optimizing by 1.5% up in mean accuracy (or AP), and furthermore the “training” improved will save 2/5 or longer of the regular working time in theory and applications. 

Several years later, this ideas—putting the pedal to the metal—might hack a new way in the fields of AI-driven technology and other relevant transformative research, which must have a say, to some extent, of the academic staff members’ potentials strenuously for epoch-making disciplinary engagements across the world, by which to prepare a community of high-quality applied talents, young researchers, as well as go-to practitioners at AXHU and beyond.