Clarification on dataset creation for MLRecogNet as SGIE to deepstream applications

Greetings,I am presently curating a dataset for the MLRecogNet classifier, as specified in the “Metric Learning Recognition: NVIDIA Docs” (
Metric Learning Recognition - NVIDIA Docs).
Where the data structure is not easily comprehensible. I require comprehensive explanations regarding the definition of a reference and a query. From the official documentation, in the section under “preparing dataset”, within the training folder there were also subfolders named as “reference” Is it critical? And what exactly is a query set.? If my comprehension is accurate, the dataset necessitates reference, validation, training, and testing. If that is the case, what distinguishes reference from validation? Must the data between those be entirely distinct?

There is no update from you for a period, assuming this is not an issue anymore. Hence we are closing this topic. If need further support, please open a new one. Thanks

Should be typo.

Yes. You can also refer to tao_tutorials/notebooks/tao_launcher_starter_kit/metric_learning_recognition/specs/train.yaml at main · NVIDIA/tao_tutorials · GitHub

During training, evaluation, and inference, MLRecogNet requires a reference set and a query set for validation or test. The reference set consists of a collection of labeled images, while the query set refers to a group of unlabeled images.

You can refer to tao_tutorials/notebooks/tao_launcher_starter_kit/metric_learning_recognition/process_retail_product_checkout_dataset.py at main · NVIDIA/tao_tutorials · GitHub and tao_tutorials/notebooks/tao_launcher_starter_kit/metric_learning_recognition/metric_learning_recognition.ipynb at main · NVIDIA/tao_tutorials · GitHub.

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.