PyTorch | Important Links & Readable
Vision aspect of PyTorch — Object & Event Detection and Response
The intent of this article is to give you a quick reference on all important links. These links will help you practice using Google Colab, review Github Code & Understand concepts.
I will just be the guide here, who has collated this information for you.
To Start with, I will help you with few steps and key concept behind PyTorch Research.
What all to Import & Why?:
By default; Import pandas, numpy, matplot.pyplot and sklearn to be on safer side.
Torch :- Top-Level PyTorch Library
Torch.nn :- PyTorch Library for Building Neural Networks
Torch.autograd :- Automatic Differentiable gradient version of PyTorch
Torch.nn.functional :- Includes Loss Function, Activation Function, etc
Torch.optim :- Optmization Operation like SGD & Adam(adaptive moment estimation)
Torch.utils :- Like Dataset and DataLoader — Data loading utility
Torchvision :- Popular dataset & pretrained models
Key Concepts
pip install:
· pip install pytorch torchvision cudatoolkit=10.0 -c pytorch
· Differentiation in PyTorch: Autograd
· Computation Model: Dynamic Computation Graph
· pip install torchviz
Optimizer:
· torch.optim.Adadelta
· torch.optim.Adagrad
· torch.optim.RMSprop
· torch.optim.Adam
Loss:
· -Mean Absolute Error: torch.nn.L1Loss
· -Mean Square Error Loss: torch.nn.MSELoss
· -Smooth L1 Loss: torch.nn.SmoothL1Loss
· -Negative Log-Likelihood Loss: torch.nn.NLLLoss
· -Cross-Entropy Loss: torch.nn.CrossEntropyLoss
· -Kullback-Leibler divergence: torch.nn.KLDivLoss
Layers:
nn.Linear,
nn.Conv2d, nn.MaxPool2d, nn.ReLU, nn.BatchNorm2d, nn.Dropout, nn.Embedding,
nn.GRU/nn.LSTM,
nn.Softmax, nn.LogSoftmax, nn.MultiheadAttention,
nn.TransformerEncoder, nn.TransformerDecoder
Important Links Now
Must One… if you already know PyTorch & Google Colab To Practice
https://colab.research.google.com/drive/1CPy4RG_tfQGm3d6UvTBEDCJZxB71VOzn?usp=sharing
https://medium.com/pytorch/nvidia-gtc-2021-hosts-over-50-pytorch-sessions-4312e9e8f3ee
Day 0:
https://github.com/pytorch/examples/blob/master/mnist/main.py
Day 1:
https://pytorch-lightning.readthedocs.io/en/latest/starter/new-project.html
https://github.com/HIPS/autograd#end-to-end-examples
https://www.cs.toronto.edu/~rgrosse/courses/csc321_2018/slides/lec10.pdf
Day 2:
https://playground.tensorflow.org/
https://cloud.google.com/vision
https://github.com/pytorch/examples
https://commons.wikimedia.org/wiki/File:Typical_cnn.png
https://commons.wikimedia.org/wiki/File:Convolutional_Neural_Network_with_Color_Image_Filter.gif
Day 3:
https://github.com/facebookresearch/SlowFast
https://pytorch.org/tutorials/beginner/deeplabv3_on_android.html
https://pytorch.org/hub/ultralytics_yolov5/
https://colab.research.google.com/github/ultralytics/yolov5/blob/master/tutorial.ipynb
https://pytorch.org/hub/intelisl_midas_v2/
https://www.kaggle.com/gokulkarthik/image-segmentation-with-unet-pytorch
Day 4:
https://www.biomotionlab.ca/html5-bml-walker/
https://www.kaggle.com/tawsifurrahman/covid19-radiography-database
CVAT Annotation Tool: CVAT Automated Annotation using your own custom models — full video
CVAT: https://cvat.org
Conclusion
Its a day-wise split so that you can keep a track on your progress & these links are primarily subjected Computer Vision aspect of PyTorch — Object & Event Detection and Response
Please clap if you have found it useful and comprehensive.