Based on this prominent work, we have taken a large collective effort to study this topic and published an improved version in the IEEE/CVF Computer Vision and Pattern Recognition Conference (CVPR), a top-tier conference in computer vision.
To overcome two crucial challenges of subcellular structure prediction (SSP, another name of label-free prediction), i.e. partial labeling and multi-scale, we propose Re-parameterizing Mixture-of-Diverse-Experts (RepMode), a network that dynamically organizes its parameters with task-aware priors to handle specified single-label prediction tasks of SSP. RepMode acheives SOTA overall performance in SSP and show its potiental in task-increment learning.
Hi guys,
Based on this prominent work, we have taken a large collective effort to study this topic and published an improved version in the IEEE/CVF Computer Vision and Pattern Recognition Conference (CVPR), a top-tier conference in computer vision.
To overcome two crucial challenges of subcellular structure prediction (SSP, another name of label-free prediction), i.e. partial labeling and multi-scale, we propose Re-parameterizing Mixture-of-Diverse-Experts (RepMode), a network that dynamically organizes its parameters with task-aware priors to handle specified single-label prediction tasks of SSP. RepMode acheives SOTA overall performance in SSP and show its potiental in task-increment learning.
See our paper and code here:
Feel free to contact me (Donghao Zhou: dh.zhou@siat.ac.cn) if anything is unclear or you are interested in potential collaboration.
Again, thanks to the authors of this repo for their excellent work!
Best regards, Donghao Zhou