Update static Vision card to native COCO training

index.html

@@ -229,14 +229,14 @@
     <a class="project" href="https://huggingface.co/AETHORIA-AI/TR-HASH-Vision-v6-1M-COCO" target="_blank" rel="noreferrer">
       <div>
         <h2>TR-HASH Vision v6</h2>
-        <p>Compact hierarchical hash-routed detector with native multi-scale features, shifted-window attention, residual level adapters and a lightweight PAN/P2 neck. The complete model is trained from scratch on Objects365 before controlled COCO refinement; no external image-classification backbone is used.</p>
+        <p>Compact hierarchical hash-routed detector with native multi-scale features, shifted-window attention, residual level adapters and a lightweight PAN/P2 neck. The complete model will be initialized randomly and trained end-to-end directly on COCO 2017 at 640 px, without external detector, classification backbone or pretrained weights.</p>
         <div class="chips">
-          <span class="chip">≈1.31M Objects365</span>
-          <span class="chip">From-scratch detector</span>
-          <span class="chip">Class × level hash gate</span>
+          <span class="chip">≈1.31M parameters</span>
+          <span class="chip">COCO from scratch</span>
+          <span class="chip">4 experts · top-2</span>
           <span class="chip">P2 + PAN + STAL</span>
           <span class="chip">DDP + Triton</span>
-          <span class="chip">Training active</span>
+          <span class="chip">Evaluation pending</span>
         </div>
       </div>
       <span class="arrow" aria-hidden="true">→</span>