multispectral-object-detection icon indicating copy to clipboard operation
multispectral-object-detection copied to clipboard

Published 20 hours ago • verified publisher icon DocF

some questions about the result

Open challow0 opened this issue 2 years ago • 5 comments

Thanks for your great work~ I find the result in LLVIP origin paper is higher than the result in your paper. I want to know why this difference happened. I did the experiment using the yolov5 v6.0 version and found it's the result in LLVIP paper. (AP 0.527 and 0.67). Can you help me to explain this question? ^_^

  • The result in LLVIP paper (using yolov5)

image

  • The result in your paper

image

challow0 avatar Apr 05 '22 06:04 challow0

Yes, the detection results on the single modality of the LLVIP dataset come from the LLVIP paper. By the way, the baseline of this repository is yolov5 v5.0 version.

DocF avatar Apr 05 '22 07:04 DocF

您好，我也想问一下就是为什么您论文中关于RGB和IR的结果跟LLVIP论文中的结果不一致呀，如果IR的结果都优于多光谱融合后的检测结果了，那做多光谱的意义是什么

yeahyeahye avatar Apr 11 '22 10:04 yeahyeahye

会不会是因为作者用的1024img_size丢了信息，LLVIP paper用的1280

Hiram1026 avatar Apr 22 '22 12:04 Hiram1026

有道理唉，我之前都没注意到这个

---Original--- From: @.> Date: Fri, Apr 22, 2022 20:48 PM To: @.>; Cc: @.@.>; Subject: Re: [DocF/multispectral-object-detection] some questions about the result (Issue #24)

会不会是因为作者用的1024img_size丢了信息，LLVIP paper用的1280

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.***>

yeahyeahye avatar Apr 22 '22 12:04 yeahyeahye

image image @DocF 这是我在yolov5上跑的红外单模态，为什么IR的指标都优于了融合后

Hiram1026 avatar Apr 26 '22 04:04 Hiram1026