Skip to content

Releases: qubvel/efficientnet

EfficientNet v1.1.1

15 Sep 16:22
8984e98
Compare
Choose a tag to compare

Fixes

  • Fix get_custom_objects() (#128)

EfficientNet v1.1.0

28 Feb 13:16
Compare
Choose a tag to compare

Updates

  • Add Noisy Student weights for b0-b7 models (weights="noisy-student")

EfficientNet - v1.0.0

04 Oct 10:08
aa1edca
Compare
Choose a tag to compare

Stable release of EfficientNet

  • keras and tf.keras support
  • imagenet pretrained weights for b0-b7 models

Efficientnet - v1.0.0b3

08 Aug 15:36
d740fc8
Compare
Choose a tag to compare
  1. Add B5 and B6 weights (thanks to @Callidior)
  2. Weights updated to better ones, trained with autoaugment (thanks to @Callidior again)

Efficientnet - v1.0.0b2

05 Aug 08:40
48e36b4
Compare
Choose a tag to compare

Fix swish custom object deserialization

Efficientnet - v1.0.0b1

02 Aug 11:42
2e5be5c
Compare
Choose a tag to compare

Improvements:

  1. Support of keras and tf.keras frameworks
  2. Preprocessing from keras_applications
  3. Reduce number of custom objects
  4. Tests

Changes

  1. API changes (model import under framework modules)

Notes

  1. Previously trained models are not compatible with current release, use model.load_weights function or roll back to previous version - v0.0.4

Efficientnet - v0.0.4

16 Jul 13:55
8aa3fa5
Compare
Choose a tag to compare

Added input_tensor param and weights for B4, B5 models.

Weights for EfficientNet

31 May 08:40
Compare
Choose a tag to compare
Pre-release
v0.0.1

Create README