search

microsoft / CameraTraps

PyTorch Wildlife: a Collaborative Deep Learning Framework for Conservation.
https://cameratraps.readthedocs.io/en/latest/
MIT License
784 stars 246 forks source link

Missing compiled model file for classification inference #352

Closed workhodge1 closed 12 months ago

workhodge1 commented 1 year ago

I'd like to run inference on my cropped data using the MegaDetector v0.1 model discussed in Step 3 here: https://github.com/microsoft/CameraTraps/tree/main/classification. However, I can not find the model file listed anywhere in the repo.

nomoneyExpection commented 1 year ago

do you find the model file now?

workhodge1 commented 1 year ago

do you find the model file now?

Not yet. Please give me a heads up if you find it.

nomoneyExpection commented 1 year ago

The result I see should be that there is no such model. This model is obtained by him based on efficientnet training. There seems to be no big difference in the detection type, and the accuracy has improved a little.

nomoneyExpection commented 1 year ago

i have found that We have not publicly released any species classification models. For some use cases, we may be able to share one or more models if you contact us at cameratraps@lila.science, but our classification models are currently targeted for very specific workflows, so we're not sharing them broadly. Others have used our classification code as a starting point for training your own classifiers that are relevant to your data; this may be a more practical starting point. Thanks.
-Dan

aa-hernandez commented 12 months ago

Thank you for your valuable input. We've moved to a new codebase with the release of Pytorch-Wildlife v1.0. The issue you mentioned might not exist anymore or it might be resolved in the new codebase. We encourage you to try the new release here and share any further feedback or concerns. If the issue still persists, please let us know through this channel and we will work on it!

If you're still interested in working with the previous version, you can take a look at Dan Morris' fork. The issue may or may not be resolved in that codebase.