Record

Release note

SMZF V1.0.0

1. This is the first version of Sunplus Model Zoo Framework.

2. SMZF V1.0.0 is a split from Model-Zoo V1.0.0, mainly responsible for the application logic part of model validation, and is independent of specific models.

3. Running smzf alone is meaningless. SMZF V1.0.0 needs to be used in conjunction with Model-Zoo V1.0.1, which has the same functional form as Model-Zoo V1.0.0. The types and quantities of models are already equivalent to SNNF V1.2.0.

Model-Zoo V1.0.1

1. Model-Zoo will only undertake the management function of the Model repository, which will include some necessary config, resources, and tools for generating models. Using it independently will have no meaning, and it will need to be used in conjunction with SMZF/SNNF.

2. Update models by VSI Acuity version 6.30.7, consider int8 or int16, hybrid or not. All the Models corresponding to SNNF1.2.0 are ready now.

Model-Zoo V1.0.0

1. Concise and convenient nbg code porting.

   1. Provide a Linux tool for model porting, which named modelCreater and can quickly convert the file information related to the model graph exported by the model toolkit into the format required by model zoo and generate the corresponding C++ code. At the same time, rename the nb file and these C++ codes together according to certain naming rules and move them to the specified path in model zoo.

   2. Model-zoo will automatically obtain the basic information of the newly added model. After adding the desired control options, the new model can be found in the demo's model list and immediately run to output the basic top 5 results.

2. Update models by VSI Acuity version 6.30.7, consider int8 or int16, hybrid or not.

   1. Only some typical models will be available now, mainly used to verify the basic application process of Model-Zoo and the combination of Model-Zoo and SNNF.

   2. All the Models corresponding to SNNF1.2.0 will be completed until Model-Zoo1.0.1.

3. Add some new NN models.

   1. Yolov11 Detection.

   2. Yolo11s OBB (oriented bounding box).

   3. Yolo11s Segmentation.

   4. Yolo11s Classificy.

   5. Yolo11s Pose.

4. Python interface for Model-Zoo user API.

   1. We will provide demo，single model mAP test sample and another more universal Python interface, temporarily named bare metal mode.

   2. About bare metal mode, only CNN hardware related with C++ code, other parts is python related, so you can use NPU resource with open source python code for your conveniently.

   3. Please get the xxx_usage.txt and xxx_test.py for your test.