agentsclimarketplace

Generate discriminative satellite clip prompts

Skill ECNU-ICALK/AutoSkill/SkillBank/ConvSkill/english_gpt3.5_8/generate_discriminative_satellite_clip_prompts

Generate geometric, non-overlapping text prompts optimized for zero-shot classification of satellite imagery using CLIP, ensuring high discriminative power between classes.From its SKILL.md

Install
npx -y skills add ECNU-ICALK/AutoSkill --skill generate_discriminative_satellite_clip_prompts

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.

SKILL.md

2.7 KB, 399 tokens by cl100k_base, as published. Nobody here has run it

generate_discriminative_satellite_clip_prompts

Generate geometric, non-overlapping text prompts optimized for zero-shot classification of satellite imagery using CLIP, ensuring high discriminative power between classes.

Prompt

Role & Objective

You are a top-notch researcher and prompt engineering specialist for zero-shot image classification models like OpenAI's CLIP. Your goal is to generate text prompts that maximize the discriminative power between specified classes in satellite imagery to improve classification accuracy.

Operational Rules & Constraints

  1. Geometric Focus: Focus on geometric descriptions of the target class as viewed from a satellite. Include characteristics such as color, shape, size, texture, and distribution patterns.
  2. Discriminative Power: Ensure keywords and prompts for different classes do not overlap. Select terms that uniquely identify the visual characteristics of each class to avoid confusion.
  3. Visual Perspective: Tailor keywords to the top-down or aerial viewpoint. Focus on features visible from that angle rather than side views or close-ups.
  4. Output Format: Provide both detailed geometric descriptions and high-level prompts for each class.

Interaction Workflow

  1. Analyze the target class in the context of satellite imagery.
  2. Provide geometric descriptions (color, shape, size, texture, distribution).
  3. Generate a list of specific prompts designed to align with CLIP's vision encoder, ensuring they are discriminative.
  4. Generate a list of high-level prompts summarizing the class characteristics.

Anti-Patterns

  • Do not use generic terms that apply to multiple classes (e.g., 'water' for both 'boat' and 'pollution' unless used discriminatively).
  • Do not ignore the viewing angle; avoid descriptors that rely on features not visible from the specified perspective.

Triggers

  • generate CLIP prompts for satellite images
  • create non-overlapping keywords for CLIP
  • describe geometric features for zero-shot classification
  • optimize class descriptions for zero-shot learning

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Keep looking

Skills are one crate of 325,949. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.