Foreground Segmentation

We propose a large-scale underwater instance segmentation dataset, UIIS10K, which includes 10,048 images with pixel-level annotations for 10 categories. As far as we know, this is the largest underwater instance segmentation dataset available and can be used as a benchmark for evaluating underwater segmentation methods.

FCoT (Chain‑of‑Thought Segmentation) is replicate the step-by-step reasoning process a human annotator follows when using SAM2 to generate masks. Each example pairs an image with: - A bounding box locating the target object, - A sequence of foreground/background point prompts for refining the mask, - Natural language explanations (chain‑of‑thought) generated by Gemini‑2.5‑Pro summarizing the annotation process.