MC Dropout for Uncertainty in Landmine and UXO Classification

Monte Carlo Dropout with a fine‑tuned ResNet‑50 estimates uncertainty for landmine and UXO image classification. Accepted for IGARSS 2025 and slated for IEEE IGARSS 2025. Read more: getnews.me/mc-dropout-for-uncertain... #montecarlodropout #resnet50

ResNet-50 with Adaptive Spatial Feature Fusion Improves Skin Lesion Classification

An ASFF‑enhanced ResNet‑50 hit 93.18% accuracy on 3,297 ISIC 2020 dermoscopic images, with AUC 0.9670 (PR) and 0.9717 (ROC); a lightweight fusion module runs on hardware. Read more: getnews.me/resnet-50-with-adaptive-... #resnet50 #asff

DAOcc boosts 3D occupancy prediction with object detection supervision

DAOcc uses a ResNet‑50 visual backbone at 256×704 resolution and achieves 54.2 % mIoU, while running at 104.9 FPS on an NVIDIA RTX 4090. Read more: getnews.me/daocc-boosts-3d-occupanc... #3doccupancy #daocc #resnet50

Scaling Decentralized Data Parallel Deep Learning with Adaptive Graphs

Ada, a decentralized adaptive graph method, achieved the best convergence rates on ResNet‑50 training on ImageNet‑1K with 1008 GPUs. Read more: getnews.me/scaling-decentralized-da... #decentralized #adaptivegraph #resnet50