OpenOOD - CIFAR10

Reproduces the OpenOOD benchmark for OOD detection, using the WideResNet model from the Hendrycks baseline paper.

Warning

This is currently incomplete, see CIFAR10-OpenOOD.

import pandas as pd  # additional dependency, used here for convenience
import torch

from pytorch_ood.benchmark import CIFAR10_OpenOOD
from pytorch_ood.detector import MaxSoftmax, ReAct, ASH
from pytorch_ood.model import WideResNet
from pytorch_ood.utils import fix_random_seed

fix_random_seed(123)

device = "cuda:0"
loader_kwargs = {"batch_size": 64}

model = WideResNet(num_classes=10, pretrained="cifar10-pt").eval().to(device)
trans = WideResNet.transform_for("cifar10-pt")
norm_std = WideResNet.norm_std_for("cifar10-pt")

Just add more detectors here if you want to test more

detectors = {
    "MSP": MaxSoftmax(model),
}

results = []
benchmark = CIFAR10_OpenOOD(root="data", transform=trans)

with torch.no_grad():
    for detector_name, detector in detectors.items():
        print(f"> Evaluating {detector_name}")
        res = benchmark.evaluate(detector, loader_kwargs=loader_kwargs, device=device)
        for r in res:
            r.update({"Detector": detector_name})
        results += res

df = pd.DataFrame(results)
print((df.set_index(["Dataset", "Detector"]) * 100).to_csv(float_format="%.2f"))

This should produce the following table:

Dataset	Detector	AUROC	AUPR-IN	AUPR-OUT	FPR95TPR
CIFAR100	MSP	87.83	85.20	88.42	43.08
TinyImageNet	MSP	87.06	85.05	86.82	51.27
MNIST	MSP	92.66	90.29	94.33	22.47
FashionMNIST	MSP	94.95	93.36	96.18	15.59
Textures	MSP	88.51	78.50	92.99	40.86
Places365	MSP	88.24	95.61	71.17	44.65