Foreground segmentation masks generated by SEEM have been added to standard datasets used for CLIP-based prompt tuning research like CoOp. The masks use RGB values of [255, 255, 255] for foreground and [0, 0, 0] for background, with the shorter image side fixed to 512 pixels. This dataset was created by author JREion and last updated on Hugging Face in April 2026.
Use Cases
- Training foreground-aware prompt tuning models based on the provided segmentation masks.
- Benchmarking CLIP adaptation techniques using datasets with explicit foreground/background separation.
- Analyzing the impact of foreground isolation on model performance for tasks mentioned in the description.
Strengths
- Includes foreground masks for all raw images, generated by the SEEM model.
- Mask generation uses a consistent RGB color scheme ([255,255,255] for foreground, [0,0,0] for background).
- Images are processed with a fixed shorter side of 512 pixels, maintaining aspect ratio.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count is unknown, which may limit suitability assessment.
- The description metadata is limited; actual data quality requires manual inspection after download.
Provenance
- Source
- Hugging Face, author JREion.
- Collection Method
- Based on original datasets used for CLIP prompt tuning, with foreground masks added.
- Time Range
- null
- Freshness
- Last updated 2026-04-03 06:44:49.
- Geography
- null