Can I use my own dataset

[SiChuChu]

Hello, thank you for your work !

I want to use SSA on custom categories and datasets. I saw you mentioned that users can customize in terms of the segmentor's architecture and the interested categories. But can I use my own datasets ? I've made some attempts. But I don't understand what is "semantic_branch_processor", I try to use one of it directly.But an error will be reported —— ValueError: You have to specify the task_input. Found None. And I guess it has something to do with the fact that I didn't set the "semantic_branch_processor" correctly. I want to know how to set up "semantic_branch_processor" if I want to use my own dataset.

Thank you very much for any reply.

Appendix command： python scripts/main_ssa.py --ckpt_path ./ckp/sam_vit_h_4b8939.pth --save_img --world_size 1 --dataset VOC2012 --data_dir /media/guo/DATA/chen/lraspp/data/VOCdevkit/VOC2012/JPEGImages --gt_path /media/guo/DATA/chen/lraspp/data/VOCdevkit/VOC2012/Annotations --out_dir output_VOC2012 Complete error reporting： Traceback (most recent call last): File "scripts/main_ssa_try.py", line 269, in main(0, args) File "scripts/main_ssa_try.py", line 248, in main semantic_segment_anything_inference(file_name, args.out_dir, rank, img=img, save_img=args.save_img, File "/media/guo/DATA/chen/SSA/Semantic-Segment-Anything/scripts/pipeline.py", line 168, in semantic_segment_anything_inference class_ids = segformer_func(img, semantic_branch_processor, semantic_branch_model, rank) File "/media/guo/DATA/chen/SSA/Semantic-Segment-Anything/scripts/segformer.py", line 5, in segformer_segmentation inputs = processor(images=image, return_tensors="pt").to(rank) File "/home/guo/anaconda3/envs/ssa/lib/python3.8/site-packages/transformers/models/oneformer/processing_oneformer.py", line 112, in call raise ValueError("You have to specify the task_input. Found None.") ValueError: You have to specify the task_input. Found None.

[OSSome01]

Were you able to figure it out?