I converted a YOLOv5 model to TFLite and integrated it into a kotlin Android app for object detection. Before deploying the TFLite model in the app, it performed exceptionally well during testing and development. However, in the app there are the following issues: The model seems to detect random objects and even draws bounding boxes in areas where there are no objects. Also One of the classes consistently has a confidence score of 100%, while the remaining classes consistently show a confidence score of 0%.
Asked
Active
Viewed 48 times
