search

SoftwareGift / FeatherNets_Face-Anti-spoofing-Attack-Detection-Challenge-CVPR2019

Code for 3rd Place Solution in Face Anti-spoofing Attack Detection Challenge @ CVPR2019,model only 0.35M!!! 1.88ms(CPU)
Other
928 stars 284 forks source link

请教一下用realsense采集的IR数据的预处理思路 #90

Closed StaringWhere closed 4 years ago

StaringWhere commented 4 years ago

作者您好，我用D415采集color,ir和depth数据，按照CASIA-SURF的论文里描述的那样，对齐后用Dlib和PRNet检测和裁切人脸。测试的结果是Depth准确率不错，但是IR的结果很不理想，开不开Laser我都试了下。看到你们也作了自己的数据集，请教一下用realsense采集的IR数据的预处理思路，感谢！ 网络：FeatherNetB_bs32-ir 模型：googledrive里的_54.pth.tar

StaringWhere commented 4 years ago

大家是怎么处理数据的阿

StaringWhere commented 4 years ago

打开auto exposure和auto white balance让结果稳定了很多，不过感觉受光线影响还是很飘

nghiabka commented 3 years ago

作者您好，我用D415采集color,ir和depth数据，按照CASIA-SURF的论文里描述的那样，对齐后用Dlib和PRNet检测和裁切人脸。测试的结果是Depth准确率不错，但是IR的结果很不理想，开不开Laser我都试了下。看到你们也作了自己的数据集，请教一下用realsense采集的IR数据的预处理思路，感谢！ 网络：FeatherNetB_bs32-ir 模型：googledrive里的_54.pth.tar

Hi , depth image and ir image in casia-surf dataset have three channels, How do you convert depth image, ir image captured from d415 to three channels?