Created by: wzl0329

模型是yolov3增强模型，使用了iouloss和louawareloss，用k-means方法根据训练集更换了anchors，anchors的大小是根据输入的图片大小608按比例计算的，没有更改其他参数，没更改前的map可以达到90，更改之后效果变差了，想问一下还可以更改哪些参数可以使模型效果变好？ YOLOv3Head: anchor_masks: [[6, 7, 8], [3, 4, 5], [0, 1, 2]] #更改了anchors anchors: [[27, 66], [39, 66], [46, 68], [44, 86], [53, 91], [59, 94], [62, 104], [77, 106], [99, 142]] norm_decay: 0. iou_aware: true iou_aware_factor: 0.4 yolo_loss: YOLOv3Loss nms: background_label: -1 keep_top_k: 100 nms_threshold: 0.45 nms_top_k: 1000 normalized: false score_threshold: 0.01 drop_block: true

YOLOv3Loss: batch_size: 16 ignore_thresh: 0.7 label_smooth: false use_fine_grained_loss: true iou_loss: IouLoss iou_aware_loss: IouAwareLoss

IouLoss: loss_weight: 2.5 max_height: 608 max_width: 608

IouAwareLoss: loss_weight: 1.0 max_height: 608 max_width: 608