Update README.md

README.md

@@ -6,11 +6,24 @@ metrics:
 ---
 CSATv2
 
-CSATv2 is a lightweight high-resolution vision backbone designed to maximize throughput at 512×512 resolution. By applying frequency-domain projection at the input stage, the model suppresses redundant spatial information and achieves extremely fast inference with only 11M parameters.
+CSATv2 is a lightweight high-resolution vision backbone designed to maximize throughput at 512×512 resolution.
+By applying frequency-domain projection at the input stage, the model suppresses redundant spatial information and achieves extremely fast inference.
+
+## Highlights
+
+- 🚀 **2,800 images/s at 512×512 resolution (A6000 1×GPU)**
+- ⚡ **Frequency-Domain Projection** for removing redundant spatial information
+- 🎯 **80.02%** ImageNet-1K Top-1 Accuracy
+- 🪶 Only **11M parameters**
+- 🧩 Suitable for **image classification** or as a **high-throughput detection backbone**
+
+This model is an improved version of the architecture used in the [paper](https://www.mdpi.com/2306-5354/10/11/1279)
+
+Special thanks to **Demino** for contributing ideas and feedback that greatly helped in lightweighting and optimizing the model.
 
 Model description
 
-![image](https://cdn-uploads.huggingface.co/production/uploads/633a801b7646c9f51a05cc92/XN6ui8uxmedZTjpVSdBKA.png)
+![image](https://cdn-uploads.huggingface.co/production/uploads/633a801b7646c9f51a05cc92/pynK0OWbjH5WUlu8L7OTj.png)
 
 This model is designed primarily for image classification tasks and can also serve as a high-throughput backbone for object detection.
 ```python