We have released Neural Network Libraries 1.4.0! InstaGAN, StarGAN and DALI iterator have been added!
Spotlight
InstaGAN
StarGAN
InstaGAN performs image-to-image translation, especially of target instances, e.g. pants to skirts, by incorporating object segmentation masks. StarGAN also performs image-to-image translation, enabling translations for multiple domains with single model.
See below the example of input images and generated translation images using each model!
| StarGAN | InstaGAN | ||
|---|---|---|---|
 |
 |
 |
 |
| Input：black hair/female | Output：blond/female | ||
 |
 |
||
| Input：black hair/male | Output：blond/female | input image (jeans) | output image (skirt) |
DALI Iterator
DALI stands for NVIDIA Data Loading Library, and enables accelerated pre-processing of input data. Note that in this version, it works with DALI <= 0.14. Please install DALI with the version specified as following.
pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/cuda/10.0 \
nvidia-dali==0.14
Layers
- Improve the speed of transpose function
- Use thrust async sort and custom make_seqeuence for faster operation.
Utilities
- Added autoforward for pycpp
- Tensorflow network Optimization for converter from onnx
- Remove non-used nnb params and enable SIGN
- Use dtype of RHS in NdArray.data setter
Bug Fix/ Documentation
- Add property decorator to all backward functions.
- Fix bug in graph engine for shared variable
- fix solver.load_states()
- Fix layer_normalization parameter shape & improve performance of normalization functions
- Update nnabla converter support status doc
- Exit with 0 when no tests in pytest
- Fix a race condition when using thrust::async::sort
- added formated code
- Add a reference to deeplabv3plus example
- Fix converter nnb-template error
