Implement ModelType interface for more torchvision models

eto-ai / rikai

Parquet-based ML data format optimized for working with unstructured data

https://rikai.readthedocs.io/en/latest/

Apache License 2.0

137 stars 19 forks source link

Implement ModelType interface for more torchvision models #551

Closed changhiskhan closed 2 years ago

changhiskhan commented 2 years ago

Currently we have SSD and fasterrcnn implemented for object detection (https://github.com/eto-ai/rikai/tree/main/python/rikai/pytorch/models)

To make it easier for pytorch users, we should consider implementing additional models from (https://pytorch.org/vision/stable/models.html). At the very least, we should have 2 or 3 image classification models, and 1 video classification model implemented.

da-liii commented 2 years ago

Hint: apply_model_spec can be used to developing ModelType:

https://github.com/eto-ai/rikai/blob/c300eeede0a19c798b70bf7466e1faf0a482a092/python/rikai/testing/utils.py#L60-L67

eddyxu commented 2 years ago

Closed via #606