Skip to content

wekaDeeplearning4j-v1.6.0
Compare
Choose a tag to compare
@basedrhys basedrhys released this 22 Jun 09:55
· 424 commits to master since this release
v1.6.0
d13e48d

This release primarily focuses on improvements and new features for the image-classification side of WekaDeeplearning4j.

Changelog

  • Pretrained Models - Add DL4J and Keras pre-trained models into the model zoo. WekaDeeplearning4j now has over 30 pre-trained models, including big names like ResNet, InceptionV3, and EfficientNet
  • Improved Feature Extraction - Enable Dl4jMlpFilter to use a pre-trained model as feature extractor, as well as use multiple feature layers, and different pooling methods
  • Add new usage examples for feature extraction and training with a pre-trained model
  • Add ImageDirectoryLoader - an easy way to load 'folder-organised' image datasets into WEKA (usage example)
  • Update Deeplearning4j backend to 1.0.0-beta7
  • Add CUDA 10.2 support
  • Drop CUDA 9.2 support

SHA256 sums

  • wekaDeeplearning4j-1.6.0.zip
    cc528b0e8bbfd869f731661a599e49cf880eac94fdba0c6d8d529c0c80ad625b
  • wekaDeeplearning4j-cuda-10.0-1.6.0-linux-x86_64.zip
    77793662457d3550e1aeebcd6d943a8dac7d3cb687dfd6d7193b7c1bf9f5c609
  • wekaDeeplearning4j-cuda-10.0-1.6.0-windows-x86_64.zip
    84e5d1459d7e00fa5664a1e35eba6fc1c05e635f4e7d3df5b793ba3f4d33976c
  • wekaDeeplearning4j-cuda-10.1-1.6.0-linux-x86_64.zip
    19fef0f8aba6ffb9cd5543712fc15a812ae7836b6b8ef8ff60c8b531c20e33e5
  • wekaDeeplearning4j-cuda-10.1-1.6.0-windows-x86_64.zip
    82aa13dd25d8b46acfa7083a498861a1fdcd3eb9e25d341a4756be626880f23f
  • wekaDeeplearning4j-cuda-10.2-1.6.0-linux-x86_64.zip
    da3a7d2b049816acdc53ac6e9738b74b118e01ae8af0c0c43fd77359052a21fa
  • wekaDeeplearning4j-cuda-10.2-1.6.0-windows-x86_64.zip
    7e02cf449fd96d92f032385659f7f46039d7c3c1aa0c402626f2fd28a9e8fd1b
Assets 13
Loading