# 光流相关

Flow基本套路

* multi scale feature pyramid
* warp and correlation，得到cost
* 对不同scale的feature pyramid计算residual flow，实现refine

## RAFT

* 先把不同scale所有可能flow的cost算好，在计算出flow后look up得到cost
* 使用GRU实现不同scale coarse to fine的refine

## FastFlowNet

> ICRA 2021, 在TX2上用TensorRt可以跑到5Hz的模型

* 在feature提取层上先用卷积+stride降采样，再用pooling降采样，得到比较好的pyramid feature，
* 在correlation层，搜索半径为3的区域全部用来计算cost，在搜索半径大于3的部分，只采样计算其中一半的元素的cost，从而减少计算量
* cost到flow的过程用的是ShuffleNet中的SBD模块
* 从小scale开始做cost2flow，warp 大scale feature后再做一次cost2flow得到residual flow，总计六次，五次是refine


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://blog.cweihang.io/ml/papers/3d_vision/flow.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.