导致其他类的loss趋近于0。解决方案是通过oversampling等方式平衡各类目标的数量。
I train this model for single class detection with my own dataset, others losses look ordinary and downtrend, only cls loss always 0. yolov3.cfg file i did modify(filter,class) this is part of my training process i put my training weights in test.py file to validate the model After ...
Hello, I trained YOLOv5_5.0and YOLOv5_6.1 version, there is only one class, why is its class loss 0? Additional No response fyy378 added the question label Nov 23, 2023 Contributor github-actions bot commented Nov 23, 2023 👋 Hello @fyy378, thank you for your interest in YOLOv5 ...
(see above for traceback): assertion failed: [] [Condition x == y did not hold element-wise:] [x (losses/fast_rcnn_cls_loss/SparseSoftmaxCrossEntropyWithLogits/Shape_1:0) = ] [101 1] [y (losses/fast_rcnn_cls_loss/SparseSoftmaxCrossEntropyWithLogits/strided_slice:0) = ] [101]...