Merge branch 'main' of github.com:zjysteven/OpenOOD into main

README.md

@@ -238,6 +238,7 @@ distance: f4d5b3 -->
 > - [x] [![gen](https://img.shields.io/badge/CVPR'23-GEN-fdd7e6?style=for-the-badge)](https://openaccess.thecvf.com/content/CVPR2023/papers/Liu_GEN_Pushing_the_Limits_of_Softmax-Based_Out-of-Distribution_Detection_CVPR_2023_paper.pdf)    ![postprocess]
 > - [x] [![nnguide](https://img.shields.io/badge/ICCV'23-NNGuide-fdd7e6?style=for-the-badge)](https://arxiv.org/abs/2309.14888)    ![postprocess]
 > - [x] [![relation](https://img.shields.io/badge/NEURIPS'23-Relation-fdd7e6?style=for-the-badge)](https://arxiv.org/abs/2301.12321)    ![postprocess]
+> - [x] [![scale](https://img.shields.io/badge/ICLR'24-Scale-fdd7e6?style=for-the-badge)](https://github.com/kai422/SCALE)    ![postprocess]
 
 > Training Methods (6):
 > - [x] [![confbranch](https://img.shields.io/badge/arXiv'18-ConfBranch-fdd7e6?style=for-the-badge)](https://github.com/uoguelph-mlrg/confidence_estimation)    ![preprocess]   ![training]
@@ -251,6 +252,7 @@ distance: f4d5b3 -->
 > - [x] [![cider](https://img.shields.io/badge/ICLR'23-CIDER-f4d5b3?style=for-the-badge)](https://github.com/deeplearning-wisc/cider)    ![training]   ![postprocess]
 > - [x] [![npos](https://img.shields.io/badge/ICLR'23-NPOS-f4d5b3?style=for-the-badge)](https://github.com/deeplearning-wisc/npos)    ![training]   ![postprocess]
 > - [x] [![t2fnorm](https://img.shields.io/badge/arXiv'23-T2FNorm-f4d5b3?style=for-the-badge)](https://arxiv.org/abs/2305.17797)    ![training]
+> - [x] [![ish](https://img.shields.io/badge/ICLR'24-ish-fdd7e6?style=for-the-badge)](https://github.com/kai422/SCALE)    ![training]
 
 
 > Training With Extra Data (3):

configs/pipelines/train/train_ish.yml

@@ -0,0 +1,44 @@
+exp_name: "'@{dataset.name}'_'@{network.name}'_'@{trainer.name}'_e'@{optimizer.num_epochs}'_lr'@{optimizer.lr}'_param'@{trainer.trainer_args.param}'_bs_'@{dataset.train.batch_size}'/s'@{seed}'"
+
+output_dir: ./results/
+save_output: True
+merge_option: default
+seed: 0
+
+num_gpus: 1
+num_workers: 8
+num_machines: 1
+machine_rank: 0
+
+preprocessor:
+ name: base
+
+pipeline:
+ name: train
+
+
+
+trainer:
+ name: ish
+ trainer_args:
+ mode: minksample_expscale
+ param: 0.85
+ layer: r1
+
+
+evaluator:
+ name: base
+
+optimizer:
+ name: sgd
+ num_epochs: 100
+ lr: 0.1
+ momentum: 0.9
+ weight_decay: 0.0005
+ weight_decay_fc: 0.00005
+ nesterov: True
+ nesterov_fc: True
+
+recorder:
+ name: base
+ save_all_models: False

configs/postprocessors/scale.yml

@@ -0,0 +1,7 @@
+postprocessor:
+ name: scale
+ APS_mode: True
+ postprocessor_args:
+ percentile: 85
+ postprocessor_sweep:
+ percentile_list: [65, 70, 75, 80, 85, 90, 95]

openood/evaluation_api/evaluator.py

@@ -12,6 +12,7 @@
 from openood.postprocessors import BasePostprocessor
 from openood.networks.ash_net import ASHNet
 from openood.networks.react_net import ReactNet
+from openood.networks.scale_net import ScaleNet
 
 from .datasets import DATA_INFO, data_setup, get_id_ood_dataloader
 from .postprocessor import get_postprocessor
@@ -113,6 +114,8 @@ def __init__(
  net = ReactNet(net)
  elif postprocessor_name == 'ash':
  net = ASHNet(net)
+ elif postprocessor_name == 'scale':
+ net = ScaleNet(net)
 
  # postprocessor setup
  postprocessor.setup(net, dataloader_dict['id'], dataloader_dict['ood'])

openood/evaluation_api/postprocessor.py

@@ -8,10 +8,10 @@
  GradNormPostprocessor, GRAMPostprocessor, KLMatchingPostprocessor, KNNPostprocessor,
  MaxLogitPostprocessor, MCDPostprocessor, MDSPostprocessor, MDSEnsemblePostprocessor,
  MOSPostprocessor, ODINPostprocessor, OpenGanPostprocessor, OpenMax, PatchcorePostprocessor,
- Rd4adPostprocessor, ReactPostprocessor, ResidualPostprocessor, SSDPostprocessor,
- TemperatureScalingPostprocessor, VIMPostprocessor, RotPredPostprocessor, RankFeatPostprocessor,
- RMDSPostprocessor, SHEPostprocessor, CIDERPostprocessor, NPOSPostprocessor, GENPostprocessor,
- NNGuidePostprocessor, RelationPostprocessor)
+ Rd4adPostprocessor, ReactPostprocessor, ResidualPostprocessor, ScalePostprocessor,
+ SSDPostprocessor, TemperatureScalingPostprocessor, VIMPostprocessor, RotPredPostprocessor, 
+ RankFeatPostprocessor, RMDSPostprocessor, SHEPostprocessor, CIDERPostprocessor, NPOSPostprocessor, 
+ GENPostprocessor, NNGuidePostprocessor, RelationPostprocessor)
 from openood.utils.config import Config, merge_configs
 
 postprocessors = {
@@ -48,6 +48,7 @@
  'opengan': OpenGanPostprocessor,
  'knn': KNNPostprocessor,
  'dice': DICEPostprocessor,
+ 'scale': ScalePostprocessor,
  'ssd': SSDPostprocessor,
  'she': SHEPostprocessor,
  'rd4ad': Rd4adPostprocessor,

openood/networks/scale_net.py

@@ -0,0 +1,49 @@
+import numpy as np
+import torch
+import torch.nn as nn
+
+
+class ScaleNet(nn.Module):
+ def __init__(self, backbone):
+ super(ScaleNet, self).__init__()
+ self.backbone = backbone
+
+ def forward(self, x, return_feature=False, return_feature_list=False):
+ try:
+ return self.backbone(x, return_feature, return_feature_list)
+ except TypeError:
+ return self.backbone(x, return_feature)
+
+ def forward_threshold(self, x, percentile):
+ _, feature = self.backbone(x, return_feature=True)
+ feature = scale(feature.view(feature.size(0), -1, 1, 1), percentile)
+ feature = feature.view(feature.size(0), -1)
+ logits_cls = self.backbone.get_fc_layer()(feature)
+ return logits_cls
+
+ def get_fc(self):
+ fc = self.backbone.fc
+ return fc.weight.cpu().detach().numpy(), fc.bias.cpu().detach().numpy()
+
+
+def scale(x, percentile=65):
+ input = x.clone()
+ assert x.dim() == 4
+ assert 0 <= percentile <= 100
+ b, c, h, w = x.shape
+
+ # calculate the sum of the input per sample
+ s1 = x.sum(dim=[1, 2, 3])
+ n = x.shape[1:].numel()
+ k = n - int(np.round(n * percentile / 100.0))
+ t = x.view((b, c * h * w))
+ v, i = torch.topk(t, k, dim=1)
+ t.zero_().scatter_(dim=1, index=i, src=v)
+
+ # calculate new sum of the input per sample after pruning
+ s2 = x.sum(dim=[1, 2, 3])
+
+ # apply sharpening
+ scale = s1 / s2
+
+ return input * torch.exp(scale[:, None, None, None])

openood/postprocessors/__init__.py

@@ -30,6 +30,7 @@
 from .react_postprocessor import ReactPostprocessor
 from .rmds_postprocessor import RMDSPostprocessor
 from .residual_postprocessor import ResidualPostprocessor
+from .scale_postprocessor import ScalePostprocessor
 from .ssd_postprocessor import SSDPostprocessor
 from .she_postprocessor import SHEPostprocessor
 from .temp_scaling_postprocessor import TemperatureScalingPostprocessor

openood/postprocessors/scale_postprocessor.py

@@ -0,0 +1,29 @@
+from typing import Any
+
+import numpy as np
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+from .base_postprocessor import BasePostprocessor
+
+
+class ScalePostprocessor(BasePostprocessor):
+ def __init__(self, config):
+ super(ScalePostprocessor, self).__init__(config)
+ self.args = self.config.postprocessor.postprocessor_args
+ self.percentile = self.args.percentile
+ self.args_dict = self.config.postprocessor.postprocessor_sweep
+
+ @torch.no_grad()
+ def postprocess(self, net: nn.Module, data: Any):
+ output = net.forward_threshold(data, self.percentile)
+ _, pred = torch.max(output, dim=1)
+ energyconf = torch.logsumexp(output.data.cpu(), dim=1)
+ return pred, energyconf
+
+ def set_hyperparam(self, hyperparam: list):
+ self.percentile = hyperparam[0]
+
+ def get_hyperparam(self):
+ return self.percentile