search

facebookresearch / CutLER

Code release for "Cut and Learn for Unsupervised Object Detection and Instance Segmentation" and "VideoCutLER: Surprisingly Simple Unsupervised Video Instance Segmentation"
Other
913 stars 90 forks source link

How to learn on other datasets #42

Closed W-hary closed 6 months ago

W-hary commented 1 year ago

Thanks for the author's contribution We noticed that the author mentioned in Features that CutLER can learn unsupervised object detectors and instance segmentors solely on ImageNet-1K. Whether CutLER can be used for our datasets, which is not included in the ImageNet-1K？

xiaoxin05 commented 8 months ago

hello，i have some problems ，can you help me ？

frank-xwang commented 6 months ago

Hi, CutLER can be used for custom datasets! You can follow the pipeline of training CutLER on ImageNet-1k:

  1. Cut: Use MaskCut to provide segmentation masks for multiple instances of each image, following this section. And place the json file under "DETECTRON2_DATASETS/YOUR-DATASET/annotations/".
  2. Learn: after getting all pseudo labels on your datasets, you can follow this section to start the unsupervised model learning process. Note: 1) If your dataset is very small, it is recommended to load our ImageNet pretrained model weights and fine-tune the model on your dataset. 2) you need to follow detectron2 to understand how to train a model on custom datasets.
  3. Self-training stage is optional.

Hope it helps.

frank-xwang commented 6 months ago

Closing this issues. Please feel free to re-open it if you meet other issues.