谢昕辰

Hi @AndreaPi Sorry for the late reply. Here is a similar issue #1391. And I'll find other references for you.

Hi @pengzhao-life Could you provide your model config? If the model is not used for binary classification, the error occurs.

Hi @Winnie202, Thanks for your error report and we appreciate it a lot. Did you make any modifications to the code or config? Is the num_classe of decode_head the same...

Hi @mengxia1994, Did you check if the pre-trained model was loaded correctly? I don't have much idea about this problem at the moment, I'll try other datasets with this model...

Hi @cspearl, Could you try to set the different thresholds to the decode_head?

Hi @fakerlove1, 请问你指的是导出 ONNX 吗？你可以使用 [MMDeploy](https://github.com/open-mmlab/mmdeploy).

Hi @tjdgus625, Could you provide your config file if it's possible? Since your training launched successfully, I think it is not a problem with the code, the problem may lie...

Hi @tjdgus6255, The num_classes of decode_head should be the same as the number of classes in your dataset.

Hi @XavierCHEN34, We have plans to support this model, but due to a lack of developers, it cannot be released for a short time. If you have implemented it, PR...

Hi @ychenghan, 训练时的pipeline会把图片裁剪到同一个尺寸，测试时如果不开启多尺度测试那么图片尺寸就是原图大小，这个并不是固定的。 在测试时我们要求传入的图片数据是 list, 因此使用训练时的pipeline会报错。 https://github.com/open-mmlab/mmsegmentation/blob/ee25adc2c1d73e22355d168fa09ae7bdac899a24/mmseg/apis/test.py#L91 https://github.com/open-mmlab/mmsegmentation/blob/master/mmseg/models/segmentors/base.py#L110 https://github.com/open-mmlab/mmsegmentation/blob/master/mmseg/models/segmentors/base.py#L74