Training CNNs in Presence of JPEG Compression: Multimedia Forensics vs Computer Vision | IEEE Conference Publication | IEEE Xplore