Merge pull request #15 from roboflow/feature-pcs-leaderboard

Update PCSClient description for improved clarity and accuracy

Author: isoceles

frontend/src/components/PCSClient.tsx

@@ -384,7 +384,7 @@ export default function PCSClient() {
       <HeroSection 
         title="Promptable Concept Segmentation"
         subtitle="Model Leaderboard"
-        description="Promptable Concept Segmentation (PCS) models identify the location of objects in an image with pixel-level precision (masks) provided text prompts, point prompts, or specific regions of an image as reference. PCS models are open vocabulary, not limited to fixed classes. Leaderboard results are shown for zero shot (AP₀ i.e. no finetuning or few shot) accuracy unless otherwise noted."
+        description="Identify the location of all instances of a visual concept in an image with pixel-level precision (or predict nothing if the concept does not exist) using text prompts, specific regions of an image as reference, or both. Model performance is measured by how well the model accurately identifies real-world objects in complex scenes."
         videoUrl="/model-leaderboard-fork/video-leaderboard-pcs.mp4"
         videoPosterUrl="/model-leaderboard-fork/video-leaderboard-pcs.avif"
       />