solov2中添加mixup

[Shanyaodedanshen]

请问solov2.yml中可以配置mixup的数据增强方法吗，我仿照yolo.yml等，添加了如下内容： with_mixup: True

• !MixupImage alpha: 1.5 beta: 1.5 ……

  drop_last: true

  use_process: true

  drop_empty: false

  mixup_epoch: 200

然而训练会报错： Traceback (most recent call last): File "C:\SOLO-P\ppdet\data\reader.py", line 46, in call data = f(data, ctx) File "C:\SOLO-P\ppdet\data\transform\batch_operators.py", line 98, in call padding_segm[:, :im_h, :im_w] = gt_segm ValueError: could not broadcast input array from shape (0) into shape (0,596,480)

C:\SOLO-P\ppdet\data\parallel_map.py:243: DeprecationWarning: The 'warn' method is deprecated, use 'warning' instead .format(sample.errmsg)) 2020-12-12 10:07:27,093-WARNING: recv endsignal from outq with errmsg[consumer[consumer-19a-1] failed to map with error:[could not broadcast input array from shape (0) in to shape (0,596,480)]] 2020-12-12 10:07:28,294-WARNING: fail to map op [RandomCrop_b25567] with error: index 6 is out of bounds for axis 0 with size 6 and stack: Traceback (most recent call last): File "C:\SOLO-P\ppdet\data\reader.py", line 46, in call data = f(data, ctx) File "C:\SOLO-P\ppdet\data\transform\operators.py", line 1987, in call sample['gt_segm'], valid_ids, axis=0) File "<__array_function__ internals>", line 6, in take File "C:\Users\OceanOptic\Anaconda3\envs\solo-p\lib\site-packages\numpy\core\fromnumeric.py", line 191, in take return _wrapfunc(a, 'take', indices, axis=axis, out=out, mode=mode) File "C:\Users\OceanOptic\Anaconda3\envs\solo-p\lib\site-packages\numpy\core\fromnumeric.py", line 58, in _wrapfunc return bound(*args, **kwds) IndexError: index 6 is out of bounds for axis 0 with size 6 希望得到帮助，谢谢。

[Shanyaodedanshen]

另外，如果想在mask label中间挖洞怎么做呢，谢谢。

[yghstill]

@Shanyaodedanshen Mixup只支持gt_box的操作，由于solov2中有gt_mask的信息，Mixup尚未支持。现在有GridMask的数据增强方法，是对图像进行'挖洞'，如果想对mask也这样做，可参考这个方法。

[Shanyaodedanshen]

@Shanyaodedanshen Mixup只支持gt_box的操作，由于solov2中有gt_mask的信息，Mixup尚未支持。现在有GridMask的数据增强方法，是对图像进行'挖洞'，如果想对mask也这样做，可参考这个方法。

好的谢谢