TY - GEN
T1 - Speed up Object Detection on Gigapixel-level Images with Patch Arrangement
AU - Fan, Jiahao
AU - Liu, Huabin
AU - Yang, Wenjie
AU - See, John
AU - Zhang, Aixin
AU - Lin, Weiyao
N1 - Funding Information:
This paper introduces a new patch arrangement frame-work for fast object detection on gigapixel-level images. Under this framework, we devise a Patch Arrangement Network (PAN), which increases the efficiency of detection by learning to arrange patches. Two arrangement modules were proposed: the first, patch filter module (PFM) selects and filters patch candidates across granularities, then a patch packing module (PPM) sequentially packs the remaining patches together into canvases. The overall frame-work is jointly optimized by policy-based reinforcement learning. The extensive experiments conducted on a gi-gapixel level image dataset PANDA highlights the benefits of our approach– an improvement of inference speed on gi-gapixel images by 5×, while maintaining an ideal performance. Acknowledgement. This paper is supported in part by the following grants: National Key Research and Development Program of China Grant (No.2018AAA0100400), National Natural Science Foundation of China (No.U21B2013, 61971277), HWUM JWS 2021 (Project AVALON). We thank the PANDA dataset group for providing the testing set and implementation details of their methods.
Publisher Copyright:
© 2022 IEEE.
PY - 2022/9/27
Y1 - 2022/9/27
N2 - With the appearance of super high-resolution (e.g., gigapixel-level) images, performing efficient object detection on such images becomes an important issue. Most ex-isting works for efficient object detection on high-resolution images focus on generating local patches where objects may exist, and then every patch is detected independently. How-ever, when the image resolution reaches gigapixel-level, they will suffer from a huge time cost for detecting numerous patches. Different from them, we devise a novel patch ar-rangement frameworkfor fast object detection on gigapixel-level images. Under this framework, a Patch Arrangement Network (PAN) is proposed to accelerate the detection by determining which patches could be packed together into a compact canvas. Specifically, PAN consists of (1) a Patch Filter Module (PFM) (2) a Patch Packing Module (PPM). PFM filters patch candidates by learning to select patches between two granularities. Subsequently, from the remaining patches, PPM determines how to pack these patches to-gether into a smaller number of canvases. Meanwhile, it generates an ideal layout of patches on canvas. These can-vases are fed to the detector to get final results. Experiments show that our method could improve the inference speed on gigapixel-level images by 5 x while maintaining great performance.
AB - With the appearance of super high-resolution (e.g., gigapixel-level) images, performing efficient object detection on such images becomes an important issue. Most ex-isting works for efficient object detection on high-resolution images focus on generating local patches where objects may exist, and then every patch is detected independently. How-ever, when the image resolution reaches gigapixel-level, they will suffer from a huge time cost for detecting numerous patches. Different from them, we devise a novel patch ar-rangement frameworkfor fast object detection on gigapixel-level images. Under this framework, a Patch Arrangement Network (PAN) is proposed to accelerate the detection by determining which patches could be packed together into a compact canvas. Specifically, PAN consists of (1) a Patch Filter Module (PFM) (2) a Patch Packing Module (PPM). PFM filters patch candidates by learning to select patches between two granularities. Subsequently, from the remaining patches, PPM determines how to pack these patches to-gether into a smaller number of canvases. Meanwhile, it generates an ideal layout of patches on canvas. These can-vases are fed to the detector to get final results. Experiments show that our method could improve the inference speed on gigapixel-level images by 5 x while maintaining great performance.
KW - categorization
KW - Efficient learning and inferences
KW - Recognition: detection
KW - retrieval
UR - http://www.scopus.com/inward/record.url?scp=85141762724&partnerID=8YFLogxK
U2 - 10.1109/CVPR52688.2022.00461
DO - 10.1109/CVPR52688.2022.00461
M3 - Conference contribution
AN - SCOPUS:85141762724
SP - 4643
EP - 4651
BT - 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
PB - IEEE
T2 - 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition
Y2 - 19 June 2022 through 24 June 2022
ER -