README.md: add PySODMetrics link

README.md

@@ -79,7 +79,7 @@ We moved away from the usual random crop approach. Instead, we designed our cust
 
 Given our focus on e-commerce, we crafted a specific test set, and in order to ease benchmarking with other models and solutions, we decided to open source part of it as the [Finegrain Product Masks Lite](https://huggingface.co/datasets/finegrain/finegrain-product-masks-lite), containing 120 pixel quality masks of common objects (both UGC and studio photos).
 
-We're using the usual metrics, namely MAE, Smeasure, Emeasure and Dice, computed with PySODMetrics. We'll add more later to account for matting aspects (transparent objects) - still a work-in-progress on our end.
+We're using the usual metrics, namely MAE, Smeasure, Emeasure and Dice, computed with [PySODMetrics](https://github.com/lartpang/PySODMetrics). We'll add more later to account for matting aspects (transparent objects) - still a work-in-progress on our end.
 
 | Model | **MAE** ↓ | **Smeasure** ↑ | **Smeasure** ↑ | **Dice** ↑ |
 |--|--|--|--|--|