Inference on custom trained model

[kaphleamrit2]

I am trying to run object detection and instance segmentation on large images using SAHI. I have trained MaskRCNN model on Detectron2 and save model as "/content/output/object_detection/model_final.pth" and config file as "/content/OD_cfg.pkl".
Then I used

detection_model = Detectron2DetectionModel(
    model_path="/content/output/object_detection/model_final.pth",
    config_path="/content/OD_cfg.pkl",
    confidence_threshold=0.5,
    image_size=640,
    device="cpu", # or 'cuda:0'
)

But I am getting the error: /content/OD_cfg.pkl not available in Model Zoo!. I can't run the prediction. I think I am missing something here. Does anybody have any idea what's going on here? How can I load custom trained model and config file? Please help me.

[fcakyon]

Thanks a lot for the issue @kaphleamrit2 !

We currently only support yaml config files for Detectron2 models. Check this post to details on how to save model config as yaml file: https://medium.com/codable/perform-sliced-inference-and-detailed-error-analysis-using-detectron2-models-37063e33073f

[kaphleamrit2]

Thank you @fcakyon. It works perfectly after I used yaml config file. Thanks for the help.

[fcakyon]

Glad to hear that @kaphleamrit2 🚀