livelossplot
tf-keras-vis
livelossplot | tf-keras-vis | |
---|---|---|
1 | 1 | |
1,282 | 306 | |
- | - | |
0.0 | 6.9 | |
almost 2 years ago | about 1 month ago | |
Python | Python | |
MIT License | MIT License |
Stars - the number of stars that a project has on GitHub. Growth - month over month growth in stars.
Activity is a relative number indicating how actively a project is being developed. Recent commits have higher weight than older ones.
For example, an activity of 9.0 indicates that a project is amongst the top 10% of the most actively developed projects that we are tracking.
livelossplot
tf-keras-vis
-
Help implementing an attention module in a DCGAN (question in comments)
Hello! I'm trying to implement the Deep convolutional GAN of this paper: Weather GAN: Multi-Domain Weather Translation Using Generative Adversarial Networks by Xuelong Li, Kai Kou, Bin Zhao (arxiv) (architecture in image). The training data the authors used consists of images and correspondingsegmentation masks with labels 0-5 for each pixel (0 being no weather-relatedpixel). I have crudely made the segmentation module G(seg), Initialgenerator module G(init) and the Discriminator D, but I don'tunderstand how to do the attention module G(att). In the paper theymentioned they used pretained weights of VGG19, but very little else is saidabout G(att).I found this https://github.com/keisen/tf-keras-vislibrary, which might help me, as I guess I would want G(att) to extractsomething like the saliency or activation maps multiplied with the image.However, I don't know what kind of layers I should use, or what practicalitiesto use apart from the input and output. Should I transfer-learn the network withthis data, and if so, with the segmentation labels (i.e. if any label 1-5 ispresent in the attention pixel returned)? Or can I use the pre-trainedimagenet?Also, does anyone know if the layer colors in thearchitecture image mean anything, or if they are just selected randomly forvisualization? Especially I'm concerned why G(att)'s last encoderlayer (3rd layer) is colored differently from the first two. I was firstthinking that maybe it means that G(att) is a module inside G(seg),but apparently not. The three middle blocks are apparently residual blocks.
What are some alternatives?
pytorch-summary - Model summary in PyTorch similar to `model.summary()` in Keras
pytorch-grad-cam - Advanced AI Explainability for computer vision. Support for CNNs, Vision Transformers, Classification, Object detection, Segmentation, Image similarity and more.
best-of-ml-python - 🏆 A ranked list of awesome machine learning Python libraries. Updated weekly.
autokeras - AutoML library for deep learning
einops - Flexible and powerful tensor operations for readable and reliable code (for pytorch, jax, TF and others)
chitra - A multi-functional library for full-stack Deep Learning. Simplifies Model Building, API development, and Model Deployment.
horovod - Distributed training framework for TensorFlow, Keras, PyTorch, and Apache MXNet. [Moved to: https://github.com/horovod/horovod]
explainable-cnn - 📦 PyTorch based visualization package for generating layer-wise explanations for CNNs.
horovod - Distributed training framework for TensorFlow, Keras, PyTorch, and Apache MXNet.
d2l-en - Interactive deep learning book with multi-framework code, math, and discussions. Adopted at 500 universities from 70 countries including Stanford, MIT, Harvard, and Cambridge.
easy_explain - A library that helps to explain AI models in a really quick & easy way