Are you trying to visualize your deep network just 2 days before the #CVPR2021 deadline? Here are some pretty alternatives to boring tensorboard graphs. (1/n)
Tensorspace: https://tensorspace.org/
Fun, interactive, 3D, and you can zoom into a specific layer for samples. (2/n)
Fun, interactive, 3D, and you can zoom into a specific layer for samples. (2/n)
PlotNeuralNet: https://github.com/HarisIqbal88/PlotNeuralNet
Pretty visualizations with direct export to tex!
https://www.overleaf.com/project/5ee51033e88684000165e37c (3/n)
Pretty visualizations with direct export to tex!
https://www.overleaf.com/project/5ee51033e88684000165e37c (3/n)
Moniel: https://github.com/mlajtos/moniel
Although discontinued, plug-n-play graph construction with simple primitives. (4/n)
Although discontinued, plug-n-play graph construction with simple primitives. (4/n)
NN SVG: http://alexlenail.me/NN-SVG/LeNet.html
Web UI for generating FCN, AlexNet, and LeNet-like architectures. (5/n)
Web UI for generating FCN, AlexNet, and LeNet-like architectures. (5/n)
Netron: https://netron.app
Neural network file viewer, (experimentally) supports most of the model formats. (6/n)
Neural network file viewer, (experimentally) supports most of the model formats. (6/n)
ENNUI: https://math.mit.edu/ennui/
This might be my favorite. You can ensemble AND train AND analyze the network online! (7/n)
This might be my favorite. You can ensemble AND train AND analyze the network online! (7/n)
There is also tensorboard ( https://pytorch.org/docs/stable/tensorboard.html), torchviz ( https://github.com/szagoruyko/pytorchviz), and hiddenlayer ( https://github.com/waleedka/hiddenlayer), all of which may be stronger on the analysis side than the visualization side. Let me know if I missed any! (8/8)