Tips, Community, and Known Issues#

Here are few general guidelines and known issues when using py-feat. We are always actively trying to improve these issues and make the toolbox easier to use. Please help us out by contributing on github!

Always spot-check your detections! While we’ve done our best to thoroughly test, benchmark, and document all the pre-trained models included in Py-Feat it’s always possible that real-world images and videos reveal a quirk or limitation of an otherwise high-performing detector.

Community#

  • Hypothesis: a social annotation tool that allows you post and read posts from other users who have visited this site. Click on the < on the top right of any page to get started.

  • Discourse Community: a Stack Overflow like forum where you can view, contribute, and vote on FAQs regarding py-feat usage. Please ask questions here first so other users can benefit from the answers!

  • Open a Github issue for all code related problems. You can do so by click the github icon on the top of any page.

Known issues#

  • Currently when performing detections and using batch_size > 1, AU models output slightly different values. This is in part due to how Py-Feat integrates the underlying detectors and we are actively working to fix this. You can follow this issue for more details.

  • Detectors can be sensitive to difference in images sizes such that very large or very small images result in very different predictions. This is largely due differences in the hyper-parameters and images sizes used to train the underlying pre-trained models. To partially help with this issue, since 0.5.0 Py-Feat supports passing keyword arguments to underlying models during initializing or detection, e.g. detector = Detector(facepose_model_kwargs={'keep_top_k': 500}). You can follow this issue for more details.