TensorNets 0.4.6

Areas of improvement

• Fix the failure when required packages was not installed (#60).

API changes

• Update init, load, and save (see examples).

Breaking changes

None.

TensorNets 0.4.5

Areas of improvement

None.

API changes

• Add init, load, and save.
• Add MobileNetv3 with pretrained weights (#58).
• Add names as parameter of middles, outputs, and weights.

Breaking changes

None.

TensorNets 0.4.3

Areas of improvement

None.

API changes

• Add EfficientNet with pretrained weights (#56).

Breaking changes

None.

TensorNets 0.4.2

Areas of improvement

• TF 2 supports (#55).
• Suppression of deprecation warnings for TF 1.14 and 1.15.
• Unit tests / CI improvements.
• Update of the performance table to follow the recent report format.

API changes

• Add reduce_max and swish.
• Add weights_regularizer as a parameter of set_args.

Breaking changes

None.

TensorNets 0.4.1

Areas of improvement

• Readability improvements.
• A bug fix for pretrained_initializer with the latest NumPy.
• Unit tests / CI improvements.
• Addition of evaluation codes for all imagenet models (see #49).

API changes

• Add logits to access a logits tensor directly.
• Add middles() which is equivalent to get_middles().
• Add outputs() which is equivalent to get_outputs().
• Add weights() which is equivalent to get_weights().
• Add summary() which is equivalent to print_summary().

Breaking changes

None.

TensorNets 0.4.0

Areas of improvement

• Reproducibility improvements.
  • Revise pretrained weight of Inception 3 (the top-5 accuracy increased by 0.038%).
  • Add missing relu in Inception2 (the top-5 accuracy increased by 0.468%).
• Bug fixes within variable_scope and name_scope (see #43).
• Unit tests / CI improvements.
• Documentation improvements.

API changes

None.

Breaking changes

None.

TensorNets 0.3.6

This is a small release that resolves installation issues on Windows.

TensorNets 0.3.5

Areas of improvement

• Reproducibility improvements.
  • Add missing relu in DenseNet (the top-5 accuracies increased by 2.58-3.2%).
  • Revise the zero paddings of pool1 in ResNets (the top-5 accuracies increased by 0.6-1.16%).
• Unit tests / CI improvements.
  • Disable tests if only README has been changed.
• New APIs: MS COCO utils.

API changes

• Add MS COCO utils.
• Add training data loader for VOC.
• Add training codes for YOLOv2.

Breaking changes

None.

TensorNets 0.3.4

Note that the following descriptions include changes for the 0.3.3. Additionally, The 0.3.2 doesn't exist due to PyPI conflicts.

Areas of improvement

• Bug fixes.
• New APIs: image classification and object detection models.
• Unit tests / CI improvements.
• Documentation improvements.
• Float16 supporting improvements (see #14).
• Python 3 compatibility improvements.

API changes

• Add image classification models (Darknet19, MobileNet35v2, ..., MobileNet140v2, and PNASNetlarge).
• Add references object detection models (YOLOv3VOC and YOLOv3COCO).
• Add operations (local_flatten and upsample).

Breaking changes

None.

TensorNets 0.3.1

This is a small release that improves dependency importing and resolves PyPI conflicts.