✨ Add GANomaly

README.md

@@ -97,6 +97,7 @@ where the currently available models are:
 - [STFPM](anomalib/models/stfpm)
 - [DFM](anomalib/models/dfm)
 - [DFKDE](anomalib/models/dfkde)
+- [GANomaly](anomalib/models/ganomaly)
 
 ## Inference
 

anomalib/models/__init__.py

@@ -47,7 +47,7 @@ def get_model(config: Union[DictConfig, ListConfig]) -> AnomalyModule:
         AnomalyModule: Anomaly Model
     """
     openvino_model_list: List[str] = ["stfpm"]
-    torch_model_list: List[str] = ["padim", "stfpm", "dfkde", "dfm", "patchcore", "cflow"]
+    torch_model_list: List[str] = ["padim", "stfpm", "dfkde", "dfm", "patchcore", "cflow", "ganomaly"]
     model: AnomalyModule
 
     if config.openvino:

anomalib/models/ganomaly/README.md

@@ -0,0 +1,30 @@
+# GANomaly: Semi-Supervised Anomaly Detection via Adversarial Training
+
+This is the implementation of the [GANomaly](https://arxiv.org/abs/1805.06725) paper.
+
+Model Type: Classification
+
+## Description
+
+GANomaly uses the conditional GAN approach to train a Generator to produce images of the normal data. This Generator consists of an encoder-decoder-encoder architecture to generate the normal images. The distance between the latent vector $z$ between the first encoder-decoder and the output vector $\hat{z}$ is minimized during training.
+
+The key idea here is that, during inference, when an anomalous image is passed through the first encoder the latent vector $z$ will not be able to capture the data correctly. This would leave to poor reconstruction $\hat{x}$ thus resulting in a very different $\hat{z}$. The difference between $z$ and $\hat{z}$ gives the anomaly score.
+
+## Architecture
+
+![GANomaly Architecture](../../../docs/source/images/ganomaly/architecture.jpg "GANomaly Architecture")
+
+## Usage
+
+`python tools/train.py --model ganomaly`
+
+## Benchmark
+
+All results gathered with seed `42`.
+
+## [MVTec Dataset](https://www.mvtec.com/company/research/datasets/mvtec-ad)
+
+### Image-Level AUC
+
+
+### Image F1 Score

anomalib/models/ganomaly/__init__.py

@@ -0,0 +1,19 @@
+"""GANomaly Model."""
+
+# Copyright (C) 2020 Intel Corporation
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions
+# and limitations under the License.
+
+from .model import GanomalyLightning
+
+__all__ = ["GanomalyLightning"]

anomalib/models/ganomaly/config.yaml

@@ -0,0 +1,120 @@
+dataset:
+  name: mvtec
+  format: mvtec
+  path: ./datasets/MVTec
+  url: ftp://guest:GU.205dldo@ftp.softronics.ch/mvtec_anomaly_detection/mvtec_anomaly_detection.tar.xz
+  category: bottle
+  task: classification
+  label_format: None
+  tiling:
+    apply: true
+    tile_size: 64
+    stride: null
+    remove_border_count: 0
+    use_random_tiling: False
+    random_tile_count: 16
+  image_size: 256
+  train_batch_size: 32
+  test_batch_size: 32
+  inference_batch_size: 32
+  num_workers: 32
+
+model:
+  name: ganomaly
+  latent_vec_size: 100
+  n_features: 64
+  extra_layers: 0
+  add_final_conv: true
+  early_stopping:
+    patience: 3
+    metric: image_AUROC
+    mode: max
+  lr: 0.0002
+  beta1: 0.5
+  beta2: 0.999
+  wadv: 1
+  wcon: 50
+  wenc: 1
+  threshold:
+    image_default: 0
+    adaptive: true
+
+project:
+  seed: 0
+  path: ./results
+  log_images_to: []
+  logger: false
+  save_to_csv: false
+
+optimization:
+  compression:
+    apply: false
+  nncf:
+    apply: false
+    input_info:
+      sample_size: null
+    compression:
+      algorithm: quantization
+      initializer:
+        range:
+          num_init_samples: 256
+    update_config:
+      init_weights: snapshot.ckpt
+      hyperparameter_search:
+        parameters:
+          lr:
+            min: 1e-4
+            max: 1e-2
+
+# PL Trainer Args. Don't add extra parameter here.
+trainer:
+  accelerator: null
+  accumulate_grad_batches: 1
+  amp_backend: native
+  amp_level: O2
+  auto_lr_find: false
+  auto_scale_batch_size: false
+  auto_select_gpus: false
+  benchmark: false
+  check_val_every_n_epoch: 2
+  checkpoint_callback: true
+  default_root_dir: null
+  deterministic: true
+  distributed_backend: null
+  fast_dev_run: false
+  flush_logs_every_n_steps: 100
+  gpus: 1
+  gradient_clip_val: 0
+  limit_predict_batches: 1.0
+  limit_test_batches: 1.0
+  limit_train_batches: 1.0
+  limit_val_batches: 1.0
+  log_every_n_steps: 50
+  log_gpu_memory: null
+  max_epochs: 100
+  max_steps: null
+  min_epochs: null
+  min_steps: null
+  move_metrics_to_cpu: false
+  multiple_trainloader_mode: max_size_cycle
+  num_nodes: 1
+  num_processes: 1
+  num_sanity_val_steps: 0
+  overfit_batches: 0.0
+  plugins: null
+  precision: 32
+  prepare_data_per_node: true
+  process_position: 0
+  profiler: null
+  progress_bar_refresh_rate: null
+  reload_dataloaders_every_epoch: false
+  replace_sampler_ddp: true
+  stochastic_weight_avg: false
+  sync_batchnorm: false
+  terminate_on_nan: false
+  tpu_cores: null
+  track_grad_norm: -1
+  truncated_bptt_steps: null
+  val_check_interval: 1.0
+  weights_save_path: null
+  weights_summary: top