Open hellock opened 3 years ago
warm up and more pretrained models
tricks like mix-up, auto-augmentation, and etc..
convert model to onnx.
EfficientNet, that could serve as backbone for mmdetection as well
some Nerual Architecture Search algorithms which could be implement on one GPU, like ENAS
multi-brach for multi-task learning
convert model to onnx.
I think pytorch 1.5 supports it. torch.onnx
darknet networks and data aug, like darknet53, mosaic, cutmix, mixup
benchmark on cifar10 and cifar100
Could you support HRNet?
Multi-label classification
More types of loss supported, such as center loss
, amsoftmax loss
, and so on.
focal loss pls!
vovnet-v2, it is faster and more accurate than resnet in my experience.
HS-ResNet: Hierarchical-Split Block on Convolutional Neural Network https://arxiv.org/abs/2010.07621 There is a reference repo: https://github.com/bobo0810/HS-ResNet, but no pretrained models are provided. It would be nice if mmcls could include this and provide some pretrained models.
The functionality of adding custom hooks like it is in mmdetection. I can take it up, if it seems like a good feature to the mmlab team.
model_parallel
AUGMIX, Stochastic Depth
TTA both of mmdet and mmseg support.
We already have timm backbone wrapper.
Ensemble Learning
We keep this issue open to collect feature requests from users and hear your voice. Our monthly release plan is also available here.
You can either: