RuntimeError: It is expected output_size equals to 2, but got size 3

RuntimeError: It is expected output_size equals to 2, but got size 3

原因：使用的是ADE20K数据集，但它的标注图片的size有三个维度，但mmsegmentation代码需要的标注图片size为两个通道的。

解决方法：在mmseg/datasets/pipelines/loading.py中找到方法LoadAnnotations，将

gt_semantic_seg=mmcv.imfrombytes(img_bytes,flag=’unchanged’,backend=self.imdecode_backend).squeeze().astype(np.uint8)

改为

gt_semantic_seg = mmcv.imfrombytes( img_bytes, flag='grayscale', backend=self.imdecode_backend).squeeze().astype(np.uint8)

改完代码就成功跑起来啦！

后面evaluation时，还出现了类似的问题

IndexError: too many indices for array: array is 2-dimensional, but 3 were indexed

File "/farm/yangweijing/code7/swin-segmentation/mmseg/core/evaluation/metrics.py", line 48, in intersect_and_union
    pred_label = pred_label[mask]

修改方法：mmseg/datasets/custom.py中get_gt_seg_maps方法里flag='unchanged’改为flag=‘grayscale’