I am seeking advice on the optimal approach for training an object detector using a dataset that may have missing objects.
Let's say my dataset, consists of images of cats and dogs. Approximately 33% have both cats and dogs labeled, 33% have only dogs labeled, and 33% have only cats labeled. Is there a strategy to utilize the 66% of the data where only one animal is labeled in order to enhance the overall accuracy of the detector? I am uncertain about how YOLO penalizes missing objects and would appreciate guidance on the best way to address this situation.
I am seeking advice on the optimal approach for training an object detector using a dataset that may have missing objects.
Let's say my dataset, consists of images of cats and dogs. Approximately 33% have both cats and dogs labeled, 33% have only dogs labeled, and 33% have only cats labeled. Is there a strategy to utilize the 66% of the data where only one animal is labeled in order to enhance the overall accuracy of the detector? I am uncertain about how YOLO penalizes missing objects and would appreciate guidance on the best way to address this situation.