Skip to main content

Convolutional Neural Networks and utilities for Computer Vision

Project description


🚧 Under construction

Convolutional Neural Networks and utilities for Computer Vision.

Models API

convnets offers implementations for the following models:

To instantiate a model you need to import the corresponding class and pass a valid configuration object to the constructor:

from convnets.models import ResNet

r18_config = {
	'l': [
		{'r': 2, 'f': 64},
		{'r': 2, 'f': 128},
		{'r': 2, 'f': 256},
		{'r': 2, 'f': 512}
	'b': False

model = ResNet(r18_config)

Or you can use one of the predefined configurations, or variants:

from convnets.models import ResNet, ResNetConfig

model = ResNet(ResNetConfig.r18)

You can find the implementation of each model and configuration examples in the convnets/models directory.

Training API

If you want to train a model in your notebooks, you can use our fit function:

form convnets.train import fit 

hist = fit(model, dataloader, optimizer, criterion, metrics, max_epochs)

You can use any Pytorch model. You will need to define the Pytorch dataloader, optimizer and criterion. For the metrics, the function expects a dict with the name of the metric as key and the metric function as value. The metric function must receive the model output and the target and return a scalar value. You can find some examples in convnets/metrics. The max_epochs parameter is the maximum number of epochs to train the model. The function will return a dict with the training history.

Additionally, we offer a training script that you can execute from the command line.

python scripts/ <path_to_config_file>

You will have to pass the path to a yaml file with the configuration for your training, including the model, optimizer, criterion, metrics, dataloader, etc. You can find some examples in the configs directory (which are timm and pytorch-lightning compatible).

We also offer Pytorch Lightning interoperability.

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

convnets-0.2.9.tar.gz (15.3 kB view hashes)

Uploaded Source

Built Distribution

convnets-0.2.9-py3-none-any.whl (21.8 kB view hashes)

Uploaded Python 3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page