: Frameworks like SimMIM showed that simple random masking strategies could help learn high-quality image representations across various architectures, including ViT and ConvNets.
Most Influential ECCV Papers (2024-09 Version) - Paper Digest Computer Vision ECCV 2022: 17th European Confe...
: This highly influential feature introduced an efficient alternative to full fine-tuning for large-scale Transformer models. By only tuning a small set of "prompts" in the input space, it allows models to adapt to new tasks with significantly lower computational costs. : Frameworks like SimMIM showed that simple random
: Features like BEVFormer used spatiotemporal transformers to learn unified BEV representations from multi-camera images, which is a critical advancement for autonomous driving perception. Computer Vision ECCV 2022: 17th European Confe...
: A simple and effective association method that improved tracking by associating almost every detection box, including those with low scores that were previously discarded. This approach significantly reduced fragmented trajectories and missing objects.