gt_classes in VOC2012.py about datasetdm

[Z1zs]

I have the same question, and I'm even more confused cause the class prediction of instance segmentation task is all "bicycle"(class 1). Have you figure it out?

[qwe3535]

I have the same question, and I'm even more confused cause the class prediction of instance segmentation task is all "bicycle"(class 1). Have you figure it out?

[Z1zs]

Hi, you can find the reply from the author here:
#15 (comment)

To summarize, this work can only be applied in category-agnostic instance segmentation settings. You may want to try similar work like Dataset Diffusion or DiffMask.

[qwe3535]

Hi, you can find the reply from the author here: #15 (comment)

To summarize, this work can only be applied in category-agnostic instance segmentation settings. You may want to try similar work like Dataset Diffusion or DiffMask.

Okay, thank you very much for your reply, it has been very helpful to me.

May I ask if you have reproduced the performance of the paper on semantic segmentation on VOC2012?

Because there are some issues with the code, I have not been able to reproduce the performance of the paper on semantic segmentation.

In VOC2012.py,
since mapper_classes = [1]
and dataset_dict["classes_str"] = [self.classes[el] for el in mapper_classes],
this means dataset_dict["classes_str"] is only [aeroplane].

In the code for generating semantic segmentation labels in train_semantic_voc.py,

outputs = seg_model(diffusion_features, controller, prompts, tokenizer, text_embeddings)；
text_embedding is generated from class_name,
class_name is dataset_dict["classes_str"],
which means class_name is [aeroplane]. Is this correct?

Shouldn't text_embedding be generated from batch["prompt"]?