Released Neural Network Libraries v1.7.0! Critical bug fix, new examples, and better models for ImageNet!

Wednesday, April 08, 2020

Release

Posted by shin

We have released Neural Network Libraries 1.7.0! prohibit_clear flag has been added to prevent unintentional removal of inplaced variable data! We have also new examples including SPADE, data augmentation, and facial keypoint detection examples! Even more, we have added better performing models for ImageNet!

Spotlight

BugFix: prohibit_clear flag to keep inplaced data from clear_buffer

We have noticed that, in previous versions, when you inplace a variable, its persistent flag was ignored, resulting in its data unintentionally getting removed. This is a bug that can lead to computation error, so we STRONGLY suggest upgrading to the latest version!

SPADE

We have implemented SPADE, which allows users to synthesize images with semantic maps and styles! Rough drawings can now be converted to photorealistic street views or beautiful scenery!

Input Output
       
       

Data augmentation examples

We have added examples showing how to implement data augmentation with various techniques including MixUp, CutMix, and VH-MixUp!

Image 1 Image 2 Mixup CutMix VH-Mixup
                   

Refer to our new data augmentation example page for further details!

Facial-keypoint-detection inference

We have implemented inference for face alignment network (FAN)!

Result by 2D-FAN Result by 3D-FAN

We can now easily detect facial landmarks, give it a try!

Better performance models and training scripts for ImageNet ILSVRC2012

We have updated models and training scripts for ImageNet ILSVRC2012! Configuration is easier now, and cosine annealing learning rate decay and mixup regularization have been added! Also, training and inference scripts for new models have been added, including RexNeXt50, SE-ResNet50 and SE-RexNeXt50!

Layers

Models

Utilities/Format Conversion

Examples

Bug Fix/Documentation​