How robust is your codebase to any custom dataset ?

tianrun-chen / SAM-Adapter-PyTorch

Adapting Meta AI's Segment Anything to Downstream Tasks with Adapters and Prompts

MIT License

1.04k stars 92 forks source link

How robust is your codebase to any custom dataset ? #80

Open SimonCoste opened 2 months ago

SimonCoste commented 2 months ago

@tianrun-chen , thanks for this very good paper and repo. I currently want to try your method and adapt it to a custom dataset ; could you point out to me how I have to format my dataset ? for example, COD10k is formatted like this, shall I follow this rule ?

├──dataset
│   ├── IMG
│   │   ├── image1.png
│   │   ├── ...
│   ├── GT
│   │   ├── image1.png
│   │   ├── ...

tianrun-chen commented 2 months ago

It is recommended to adhere to the guidelines, allowing you to make minor adjustments to the YAML file and dataloader. However, you can always modify the dataloader to accommodate different dataset structures.