Keypoint detection is the task of identifying keypoints in images and their associated classes.
The ~flash.image.keypoint_detection.model.KeypointDetector
and ~flash.image.keypoint_detection.data.KeypointDetectionData
classes internally rely on IceVision.
Let's look at keypoint detection with BIWI Sample Keypoints (center of face) from IceData. Once we've downloaded the data, we can create the ~flash.image.keypoint_detection.data.KeypointDetectionData
. We select a keypoint_rcnn
with a resnet18_fpn
backbone to use for our ~flash.image.keypoint_detection.model.KeypointDetector
and fine-tune on the BIWI data. We then use the trained ~flash.image.keypoint_detection.model.KeypointDetector
for inference. Finally, we save the model. Here's the full example:
../../../flash_examples/keypoint_detection.py
The keypoint detector can be used directly from the command line with zero code using flash_zero
. You can run the above example with:
flash keypoint_detection
To view configuration options and options for running the keypoint detector with your own data, use:
flash keypoint_detection --help