Fsoft-AIC / LGD

Dataset and Code for CVPR 2024 paper "Language-driven Grasp Detection."
https://airvlab.github.io/grasp-anything/
MIT License
22 stars 2 forks source link

Dataset #6

Closed xwccchong closed 1 month ago

xwccchong commented 1 month ago

Hello! Thank you for your excellent work! When I was doing inference, the matching of dataset names confused me. Does file_path refer to the content in the original GraspAnything dataset, and does add_file_path refer to the additional content added in GraspAnything++? At the same time, I didn't see any dataset about grasp prompts in the file list of huggingface. Can you provide a detailed description of the dataset file?

code

The file path is LGD/utils/data/grasp_anywhere_data.py

    self.grasp_files = glob.glob(os.path.join(addition_file_path, 'positive_grasp', '*.pt'))
    self.prompt_files = glob.glob(os.path.join(file_path, 'prompt', '*.pkl'))
    self.prompt_dir = os.path.join(file_path, 'prompt')
    self.instruction_dir = os.path.join(addition_file_path, 'grasp_instructions')
    self.rgb_files = glob.glob(os.path.join(file_path, 'image', '*.jpg'))
    self.rgb_dir = os.path.join(file_path, 'image')
    self.mask_files = glob.glob(os.path.join(file_path, 'mask', '*.npy'))

GraspAnthing GraspAnything pp

LogSSim commented 6 days ago

@xwccchong So, what is add_file_path? Could you share some details about this?