search

UX-Decoder / Semantic-SAM

[ECCV 2024] Official implementation of the paper "Semantic-SAM: Segment and Recognize Anything at Any Granularity"
2.39k stars 118 forks source link

How to run the reproduction training of SAM? #78

Open Cbtor opened 1 year ago

Cbtor commented 1 year ago

I tried the command python train_net.py --resume --num-gpus $n --config-file configs/semantic_sam_reproduce_sam_swinL.yaml COCO.TEST.BATCH_SIZE_TOTAL=$n SAM.TEST.BATCH_SIZE_TOTAL=$n SAM.TRAIN.BATCH_SIZE_TOTAL=$n but found that there are some training configs missing such as MIN_SIZE_TRAIN and CROP.TYPE. After I add these configs, there are still some errors, like

  File "/home/ma-user/work/detectron2-xyz-main/detectron2/data/dataset_mapper.py", line 154, in __call__
    image = utils.read_image(dataset_dict["file_name"], format=self.image_format)
KeyError: 'file_name'

when I use the sam json mapper.

Could you please give a detailed guide to show how to run this training experiment?

woshitiancai321 commented 1 year ago

do you solve this problem?

Cbtor commented 1 year ago

do you solve this problem?

Not yet @woshitiancai321

Suzi1213 commented 11 months ago

do you solve this problem?

Not yet @woshitiancai321 Where did you download the data set? Thank you for your answer