|
|
---
|
|
|
license: mit
|
|
|
---
|
|
|
|
|
|
# DetCon PyTorch
|
|
|
|
|
|
PyTorch implementation of DetCon-B from ["Efficient Visual Pretraining with Contrastive Detection"](https://arxiv.org/abs/2103.10957) (Henaff et al., ICCV 2021).
|
|
|
|
|
|
## Pretrained Weights
|
|
|
|
|
|
We provide DetCon-B ResNet-50 weights pretrained on ImageNet-1K for 300 epochs:
|
|
|
|
|
|
| Format | Download | Use Case |
|
|
|
|--------|----------|----------|
|
|
|
| Original | detconb_resnet50_imagenet300.pth | Direct loading |
|
|
|
| Torchvision | detconb_resnet50_imagenet300_torchvision.pth | MMSegmentation |
|
|
|
| Detectron2 | detconb_resnet50_imagenet300_d2.pkl | Detectron2 |
|
|
|
|
|
|
## Usage
|
|
|
|
|
|
See [Github](https://github.com/KKallidromitis/detcon-pytorch) repo for how to use weights
|
|
|
|
