detection 我是用自己数据训练 就是那个官方balloon数据，典型小样本，效果太差

[BoFan-tunning]

optimizer = dict(type='SGD', lr=0.05, momentum=0.9, weight_decay=0.0001)
optimizer_config = dict(grad_clip=None)

learning policy

lr_config = dict(
policy='step',
warmup='linear',
warmup_iters=500,
warmup_ratio=0.001,
step=[8, 11])
runner = dict(type='EpochBasedRunner', max_epochs=20)

[BoFan-tunning]

2024-02-02 17:14:29,451 - mmdet - INFO -
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.031
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=1000 ] = 0.095
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=1000 ] = 0.000
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=1000 ] = 0.000
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=1000 ] = 0.000
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=1000 ] = 0.039
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.104
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=300 ] = 0.104
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=1000 ] = 0.104
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=1000 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=1000 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=1000 ] = 0.144

2024-02-02 17:14:29,451 - mmdet - INFO -
+----------+-------+
| category | AP |
+----------+-------+
| balloon | 0.031 |
+----------+-------+
2024-02-02 17:14:29,497 - mmdet - INFO - The previous best checkpoint F:\OpenGVLab\InternImage\detection\work_dirs\mask_rcnn_internimage_t_fpn_1x_coco\best_bbox_mAP_epoch_17.pth was removed
2024-02-02 17:14:31,301 - mmdet - INFO - Now best checkpoint is saved as best_bbox_mAP_epoch_20.pth.
2024-02-02 17:14:31,301 - mmdet - INFO - Best bbox_mAP is 0.0313 at 20 epoch.
2024-02-02 17:14:31,301 - mmdet - INFO - Exp name: mask_rcnn_internimage_t_fpn_1x_coco.py
2024-02-02 17:14:31,301 - mmdet - INFO - Epoch(val) [20][13] bbox_mAP: 0.0313, bbox_mAP_50: 0.0958, bbox_mAP_75: 0.0033, bbox_mAP_s: 0.0000, bbox_mAP_m: 0.0000, bbox_mAP_l: 0.0393, bbox_mAP_copypaste: 0.0313 0.0958 0.0033 0.0000 0.0000 0.0393, segm_mAP: 0.0306, segm_mAP_50: 0.0947, segm_mAP_75: 0.0000, segm_mAP_s: 0.0000, segm_mAP_m: 0.0000, segm_mAP_l: 0.0393, segm_mAP_copypaste: 0.0306 0.0947 0.0000 0.0000 0.0000 0.0393

[BoFan-tunning]

20 训练乱好惨，怎么提供效果，只能加样本或者是训练轮次吗？感觉上这模型对数据部明个