I hope this message finds you well.
I am currently working on developing a model aimed at detecting keyboards. My objective extends beyond merely identifying a bounding box around the entire keyboard; I wish to detect each individual key and infer the keyboard layout, such as ANSI or ISO. To achieve this, I attempted to train a pose model using a class designated as “ANSI.” However, the model appears to have focused solely on bounding box detection, neglecting the keypoints entirely. Consequently, the metrics for pose detection remain at zero, and the pose loss does not exhibit any changes throughout the training process.
I appreciate any insights or recommendations you may have regarding this challenge.
Best regards
Project Type: Keypoint Detection
Operating System: Linux Mint 22.1 Cinnamon
Browser: Mozilla Firefox 134.0.2 (64-bit)
Project ID: keyboard-mk1av
Project Universe Link: keyboard Keypoint Detection Dataset and Pre-Trained Model by Aria