Logging¶

Anomalib offers various mechanisms for logging metrics and predicted masks.

Enabling Logging¶

These can be enabled using the logger parameter in logging section of each model configuration file. The available options are comet, tensorboard, wandb and csv. For example, to log to TensorBoard:

logging:
    logger: "tensorboard"

You can also pass a list of loggers to enable multiple loggers. For example:

logging:
  logger: [comet, tensorboard, wandb]
  log_graph: false

Logging Images¶

Anomalib allows you to save predictions to the file system by setting log_images: True in the visualization section . As of the current version, Anomalib also supports Comet, TensorBoard and Weights and Biases loggers for logging images. These loggers extend upon the base loggers by providing a common interface for logging images. You can access the required logger from trainer.loggers. Then you can use logger.add_image method to log images. For a complete overview of this method refer to our API documentation.

visualization:
  show_images: False # show images on the screen
  save_images: False # save images to the file system
  log_images: True # log images to the available loggers (if any)
  image_save_path: null # path to which images will be saved
  mode: full # options: ["full", "simple"]

logging:
  logger: [comet, tensorboard, wandb]
  log_graph: false

Note

Logging images to Comet,TensorBoard and wandb won’t work if you don’t have logger: [comet, tensorboard, wandb] set as well. This ensures that the respective logger is passed to the trainer object.

comet dashboard showing logged images — Comet Images in TensorBoard Dashboard¶

tensorboard dashboard showing logged images — Logged Images in TensorBoard Dashboard¶

wandb dashboard showing logged images — Logged Images in wandb Dashboard¶

Logging Other Artifacts¶

To log other artifacts to the logger, you can directly access the logger object and call its respective method. Some of the examples mentioned here might require making changes to parts of anomalib outside the lightning model such as train.py.

Note

When accessing the base logger/logger.experiment object, refer to the documentation of the respective logger for the list of available methods.

Anomalib makes it easier to log your model graph to Comet, TensorBoard or Weights and Biases. Just set log_graph to True under logging parameter of the model configuration file.

logging:
  logger: [comet, tensorboard]
  log_graph: true

comet dashboard showing model graph — Model Graph in Comet Dashboard¶

tensorboard dashboard showing model graph — Model Graph in TensorBoard Dashboard¶