AgML is a centralized framework for agricultural machine learning. AgML provides access to public agricultural datasets for common agricultural deep learning tasks, with standard benchmarks and pretrained models, as well the ability to generate synthetic data and annotations.
Apache License 2.0
167
stars
28
forks
source link
Introduce benchmarks and pretrained models + custom datasets #25
This is the precursor for AgML v0.3.0, introducing the agml.models module. The main new features introduced in this release are:
agml.models module:
Model architectures for image classification, semantic segmentation, and object detection.
Benchmarks and pretrained weights for semantic segmentation and object detection models.
agml.models.metrics module with image classification accuracy and object detection mean average precision.
agml.data updates:
Load multiple datasets into a single AgMLDataLoader.
Load custom datasets using agml.data.AgMLDataLoader.custom().
Better functionality for reducing datasets: take a random quantity of samples with take_random or take all samples with a specific class with take_class.
Other major feature improvements include:
Transforms are now applied in an order based on the time at which they are called for the AgMLDataLoader, rather than being ordered by category.
New methods in the visualization module for better and easier displaying of images and results.
A wide range of bugfixes (to be detailed further in the official release).
This is the precursor for AgML v0.3.0, introducing the
agml.models
module. The main new features introduced in this release are:agml.models
module:agml.models.metrics
module with image classification accuracy and object detection mean average precision.agml.data
updates:agml.data.AgMLDataLoader.custom()
.take_random
or take all samples with a specific class withtake_class
.Other major feature improvements include:
AgMLDataLoader
, rather than being ordered by category.