add visual_recognition/LSNet笔记

visual_recognition/LSNet Location-Sensitive Visual Recognition with Cross-IOU Loss笔记.md

@@ -0,0 +1,37 @@
+# Location-Sensitive Visual Recognition with Cross-IOU Loss笔记
+
++ Paper: [Location-Sensitive Visual Recognition with Cross-IOU Loss](https://arxiv.org/abs/2104.04899)
++ Code: [Duankaiwen/LSNet](https://github.com/Duankaiwen/LSNet)
+
+## 0.Summary Keywords
+
++ unified framework (not multi-task unified model)
++ without using the heatmaps
+
+## 1. Introduction
+
+### 1.1 Why
+
+the bounding boxes locate objects
+simply and efficiently but lack the details  
+
+masks and
+keypoints reflect the shape and pose of the objects but usually need the bounding boxes to locate object firstly  
+
+### 1.2 what
+
++ LSNet (unifies three location-sensitive visual recognition tasks)
++ cross-IOU loss (is friendly to receiving supervision from multiple scale)
++ a pyramid of deformable convolution (extracts discriminative visual cues around the landmarks)
+
+### 1.3 How
+
+### 1.4 Contributions
+
++ First, we present the formulation of location-sensitive visual recognition that inspires the community to consider the common property of these tasks
++ Second, we propose the LSNet as a unified framework in which the key technical contribution is the cross-IOU loss
+
+
+
+
+