A library for video frame prediction using SimVP, PredRNN++, MIM, PredNet, Novel GAN and Causal LSTM.
Project description
Predicto
Predicto is a Python library for video frame prediction, featuring three state-of-the-art models: PredRNN++, MIM, and Causal LSTM. This library is designed to cater to both expert and non-expert users, providing an API for developers and a simple interface for non-experts.
<h2>Features</h2>
<ul>
<li>Three video frame prediction models: PredRNN++, MIM, and Causal LSTM.</li>
<li>Easy-to-use interface for training and testing models.</li>
<li>Supports custom dataloaders or default to MovingMNIST dataset.</li>
<li>Pre and post-processing for input and output in each model.</li>
</ul>
<h2>Installation</h2>
<pre><code>pip install predicto</code></pre>
<h2>Usage</h2>
<h3>Quick Start</h3>
<pre><code>from predicto import PredRNN, MIM, ConvLSTM, SimVP,PredNet, Novel GAN and Predicto
Create a model object
model_object = MIM()
Initialize Predicto with the model object
model = Predicto(model_object)
Train the model
model.train(train_loader)
Test the model
model.test(test_loader)
<h3>Custom Dataloader</h3>
<pre><code>from predicto import PredRNN, MIM, CausalLSTM, Predicto
Define your custom dataloader
class CustomDataLoader: def init(self, ...): ...
def __iter__(self):
...
Create a model object
model_object = CausalLSTM()
Initialize Predicto with the model object and custom dataloader
model = Predicto(model_object, dataloader=CustomDataLoader())
Train the model
model.train(train_loader)
Test the model
model.test(test_loader)
<h2>Models</h2>
<ul>
<li><strong>PredRNN++</strong>: A recurrent neural network model for video frame prediction.</li>
<li><strong>MIM</strong>: Memory In Memory network for spatiotemporal predictive learning.</li>
<li><strong>Causal LSTM</strong>: A causal LSTM model for video frame prediction.</li>
</ul>
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for vpredicto-0.2.1-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 17f901633019234aae528f6eb04e792bd52580ac2163b2a09854d63913e076a4 |
|
MD5 | b38e018bd180be0ecc5a74dca3ac318d |
|
BLAKE2b-256 | 89c80da03c95b3fcb8dacfb9fb3302ef0a88966a8e6d6cb861787dfd4bc7ae16 |