请教关于增加掩码缓存的可能性
- Dominant language
- Python
- Stars
- 11.5k
- Forks
- 877
- PR merge metrics
- No merged PRs in 30d
Description
目前似乎每次都是实时推理,于是在很多场景可行性堪忧,成本敏感的场景接受不了这种成本,时间敏感的场景接受不了这种速度。
有没有可能在推理时增加缓存,通过某个 Agent 定义设备 GUI 中不变的部分,提供掩码,用图像算法如 pHash 来做掩码缓存,实现针对任意一个最原子的操作,能够通过这种高速/非大模型方法来判断设备是否处于预期的状态,如果符合记忆就直接操作,操作之后可以用完全类似的方案来判断设备是否继续符合预期。
Agent 针对未命中缓存的情况,可能是 App 版本更新,可以再引入记忆更新的方法。
Contributor guide
No contributing guide indexed for this repository
Research direction
The issue names no files, tests, or entry points. Start by tracing the inference and Agent-definition paths, then assess how GUI masks, pHash matching, cache misses, and memory updates would fit together. Done would require an agreed design and validation that cached state checks safely reduce repeated model inference without breaking actions.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- ai, computer-vision, performance
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 20/100