A5083779629- Cell Image Analysis Techniques
- Advanced Neural Network Applications
- Multimodal Machine Learning Applications
- Human Pose and Action Recognition
- Advanced Electron Microscopy Techniques and Applications
- Force Microscopy Techniques and Applications
- Advanced Fluorescence Microscopy Techniques
- Neural dynamics and brain function
- Neural Networks and Applications
- Image Processing Techniques and Applications
- Advanced Neuroimaging Techniques and Applications
- Domain Adaptation and Few-Shot Learning
- Advanced Memory and Neural Computing
- Functional Brain Connectivity Studies
- Digital Imaging for Blood Diseases
- Neuroscience and Neural Engineering
Harvard University Press
2021
Harvard University
2017-2021
Connectomics has recently begun to image brain tissue at nanometer resolution, which produces petabytes of data. This data must be aligned, labeled, proofread, and formed into graphs, each step this process requires visualization for human verification. As such, we present the BUTTERFLY middleware, a scalable platform that can handle massive interactive in connectomics. Our outputs geometry suitable hardware-accelerated rendering, abstracts low-level wrangling enable faster development new...
For video and volumetric data understanding, 3D convolution layers are widely used in deep learning, however, at the cost of increasing computation training time. Recent works seek to replace layer with blocks, e.g. structured combinations 2D 1D layers. In this paper, we propose a novel block, Parallel Separable Convolution (PmSCn), which applies m parallel streams n one along different dimensions. We first mathematically justify need (Pm) single through tensor decomposition. Then jointly...
Abstract A connectivity graph of neurons at the resolution single synapses provides scientists with a tool for understanding nervous system in health and disease. Recent advances automatic image segmentation synapse prediction electron microscopy (EM) datasets brain have made reconstructions possible nanometer scale. However, sometimes struggles to segment large correctly, requiring human effort proofread its output. General proofreading involves inspecting volumes correct errors pixel...
We present an interactive approach to train a deep neural network pixel classifier for the segmentation of neuronal structures. An training scheme reduces extremely tedious manual annotation task that is typically required networks perform well on image problems. Our proposed method employs feedback loop captures sparse annotations using graphical user interface, trains based recent and past annotations, displays prediction output users in almost real-time. implementation algorithm also...
We present a recurrent network for 3D reconstruction of neurons that sequentially generates binary masks every object in an image with spatio-temporal consistency. Our models consistency two parts: (i) local, which allows exploring non-occluding and temporally-adjacent relationships bi-directional recurrence. (ii) non-local, long-range the temporal domain skip connections. proposed is end-to-end trainable from input to sequence masks, and, compared methods relying on boundaries, its output...
A connectivity graph of neurons at the resolution single synapses provides scientists with a tool for understanding nervous system in health and disease. Recent advances automatic image segmentation synapse prediction electron microscopy (EM) datasets brain have made reconstructions possible nanometer scale. However, sometimes struggles to segment large correctly, requiring human effort proofread its output. General proofreading involves inspecting volumes correct errors pixel level,...
We present an interactive approach to train a deep neural network pixel classifier for the segmentation of neuronal structures. An training scheme reduces extremely tedious manual annotation task that is typically required networks perform well on image problems. Our proposed method employs feedback loop captures sparse annotations using graphical user interface, trains based recent and past annotations, displays prediction output users in almost real-time. implementation algorithm also...
We present a recurrent network for the 3D reconstruction of neurons that sequentially generates binary masks every object in an image with spatio-temporal consistency. Our models consistency two parts: (i) local, which allows exploring non-occluding and temporally-adjacent relationships bi-directional recurrence. (ii) non-local, long-range temporal domain skip connections. proposed is end-to-end trainable from input to sequence masks, and, compared methods relying on boundaries, its output...