[Bug]: Trained EfficientAd network can't detect any anomaly

[westwind1027]

Describe the bug

I trained an EfficientAd network by following the instruction of "https://anomalib.readthedocs.io/en/latest/markdown/get_started/anomalib.html"

I only changed “train_batch_size” to 1 in training code.

Training code

from anomalib.data import MVTec
from anomalib.engine import Engine
from anomalib.models import EfficientAd

datamodule = MVTec(
root="./datasets/MVTec", # Path to download/store the dataset
category="bottle", # MVTec category to use
train_batch_size=1, # Number of images per training batch
eval_batch_size=32, # Number of images per validation/test batch
num_workers=8, # Number of parallel processes for data loading
)

model = EfficientAd()

engine = Engine(max_epochs=10) # Train for 10 epochs

engine.fit(datamodule=datamodule, model=model)

After training finished, I got a trained "model.ckpt" in results folder. Then, I want to check if it can work as expected. So, I followed the instruction of "Lightning Inference" to inference an anomaly bottle image which belongs to MVTec dataset. I changed "dataset path" to an anomaly bottle image path, and changed "ckpt_path" to my trained "model.ckpt" in lightning inference code.

But I got an empty list "predictions" when I ran the inference code. No anomaly were found. I don't know why the trained "model.ckpt" can't work.

Lightning inference code

from pathlib import Path

from anomalib.data import PredictDataset
from anomalib.engine import Engine
from anomalib.models import EfficientAd

model = EfficientAd()
engine = Engine()

dataset = PredictDataset(
path=Path("path/to/test/images"),
image_size=(256, 256),
)

predictions = engine.predict(
model=model,
dataset=dataset,
ckpt_path="path/to/model.ckpt",
)

if predictions is not None:
for prediction in predictions:
image_path = prediction.image_path
anomaly_map = prediction.anomaly_map # Pixel-level anomaly heatmap
pred_label = prediction.pred_label # Image-level label (0: normal, 1: anomalous)
pred_score = prediction.pred_score # Image-level anomaly score

Dataset

MVTec

Model

Other (please specify in the field below)

Steps to reproduce the behavior

1. install anomalib with: pip install anomalib==2.0.0b2
2. run training code as the instruction of "https://anomalib.readthedocs.io/en/latest/markdown/get_started/anomalib.html".
3. run lightning inference code as the instruction of "https://anomalib.readthedocs.io/en/latest/markdown/get_started/anomalib.html".
4. get an empty "predictions" when inference an anomaly bottle image.

OS information

OS information:

• OS: win11 pro
• Python version: 3.11.9
• Anomalib version: 2.0.0b2
• PyTorch version: 2.3.1+cu121
• CUDA/cuDNN version: 12.6
• GPU models and configuration: 1x GeForce RTX 4080

Expected behavior

There should be some anomaly data in "predictions".

Screenshots

No response

Pip/GitHub

pip

What version/branch did you use?

2.0.0b2

Configuration YAML

default yaml configuration.

Logs

No logs.

Code of Conduct

• I agree to follow this project's Code of Conduct

[alexriedel1]

Whats the length of your dataset: len(dataset)?

[westwind1027]

Whats the length of your dataset: len(dataset)?

len(dataset) is 1.