Merged PR 1163: Gallery updates for Pi3 specific ELL models

Make the following Gallery updates: - Update ms/frame and info pages for Pi3 optimized models - Landing pages for optimized models now include a xxxx_pi.ell.zip - Update all_models.json to reflect new ms/frame and to remove Buckthorn and CandleNut since they are no longer frontier models - Address debt item to change from s/frame to ms/frame in model info pages NOTE: This change is dependent on microsoft/ELL-models#70 in the ELL-Models repo, as well as all the optimizer PRs.
microsoft · Feb 28, 2019 · 158d62a · 158d62a
1 parent 0aa70e7
commit 158d62a
Show file tree

Hide file tree

Showing 82 changed files with 220 additions and 238 deletions.
diff --git a/docs/gallery/ILSVRC2012/Bean.md b/docs/gallery/ILSVRC2012/Bean.md
@@ -6,20 +6,20 @@ permalink: /gallery/ILSVRC2012/Bean
 
 [Back to Gallery](/ELL/gallery)
 
-## ILSVRC2012 Classification: 128x128x3 Convolutional Neural Network (55.12% top 1 accuracy, 78.21% top 5 accuracy, 0.145s/frame on Raspberry Pi 3 (Raspbian) @ 700MHz)
+## ILSVRC2012 Classification: 128x128x3 Convolutional Neural Network (55.12% top 1 accuracy, 78.21% top 5 accuracy, 144ms/frame on Raspberry Pi 3 (Raspbian) @ 700MHz)
 
 <table class="table table-striped table-bordered">
     <tr>
         <td> Download </td>
-        <td colspan="3"> <a href="https://github.com/Microsoft/ELL-models/raw/master/models/ILSVRC2012/Bean/Bean.ell.zip">Bean.ell.zip</a></td>
+        <td> <a href="https://github.com/Microsoft/ELL-models/raw/master/models/ILSVRC2012/Bean/Bean_pi3.ell.zip">Bean_pi3.ell.zip</a></td>
     </tr>
     <tr>
         <td> Accuracy </td>
         <td colspan="3"> ILSVRC2012: 78.21% (Top 5), 55.12% (Top 1) </td>
     </tr>
     <tr>
         <td> Performance </td>
-        <td colspan="3"> Raspberry Pi 3 (Raspbian) @ 700MHz: 0.145s/frame </td>
+        <td colspan="3"> Raspberry Pi 3 (Raspbian) @ 700MHz : 144ms/frame </td>
     </tr>
     <tr>
         <td> Uncompressed Size </td>
@@ -41,7 +41,7 @@ permalink: /gallery/ILSVRC2012/Bean
     </tr>
     <tr>
         <td> Notes </td>
-        <td colspan="3"> Trained by Juan Lema using PyTorch 0.4.0 </td>
+        <td colspan="3"> Trained by Byron Changuion using PyTorch 1.0.0 </td>
     </tr>
 </table>
 
diff --git a/docs/gallery/ILSVRC2012/BrazilianGuava.md b/docs/gallery/ILSVRC2012/BrazilianGuava.md
@@ -6,7 +6,7 @@ permalink: /gallery/ILSVRC2012/BrazilianGuava
 
 [Back to Gallery](/ELL/gallery)
 
-## ILSVRC2012 Classification: 64x64x3 Convolutional Neural Network (44.01% top 1 accuracy, 68.64% top 5 accuracy, 0.08s/frame on Raspberry Pi 3)
+## ILSVRC2012 Classification: 64x64x3 Convolutional Neural Network (44.01% top 1 accuracy, 68.64% top 5 accuracy, 80ms/frame on Raspberry Pi 3)
 
 <table class="table table-striped table-bordered">
     <tr>
@@ -19,7 +19,7 @@ permalink: /gallery/ILSVRC2012/BrazilianGuava
     </tr>
     <tr>
         <td> Performance </td>
-        <td colspan="3"> Raspberry Pi 3 (Raspbian) @ 700MHz: 0.08s/frame </td>
+        <td colspan="3"> Raspberry Pi 3 (Raspbian) @ 700MHz : 80ms/frame </td>
     </tr>
     <tr>
         <td> Uncompressed Size </td>

diff --git a/docs/gallery/ILSVRC2012/Buckthorn.md b/docs/gallery/ILSVRC2012/Buckthorn.md
@@ -6,20 +6,20 @@ permalink: /gallery/ILSVRC2012/Buckthorn
 
 [Back to Gallery](/ELL/gallery)
 
-## ILSVRC2012 Classification: 64x64x3 Convolutional Neural Network (57.57% top 1 accuracy, 80.55% top 5 accuracy, 0.192s/frame on Raspberry Pi 3 (Raspbian) @ 700MHz)
+## ILSVRC2012 Classification: 64x64x3 Convolutional Neural Network (57.57% top 1 accuracy, 80.55% top 5 accuracy, 171ms/frame on Raspberry Pi 3 (Raspbian) @ 700MHz)
 
 <table class="table table-striped table-bordered">
     <tr>
         <td> Download </td>
-        <td colspan="3"> <a href="https://github.com/Microsoft/ELL-models/raw/master/models/ILSVRC2012/Buckthorn/Buckthorn.ell.zip">Buckthorn.ell.zip</a></td>
+        <td colspan="3"> <a href="https://github.com/Microsoft/ELL-models/raw/master/models/ILSVRC2012/Buckthorn/Buckthorn_pi3.ell.zip">Buckthorn_pi3.ell.zip</a></td>
     </tr>
     <tr>
         <td> Accuracy </td>
         <td colspan="3"> ILSVRC2012: 80.55% (Top 5), 57.57% (Top 1) </td>
     </tr>
     <tr>
         <td> Performance </td>
-        <td colspan="3"> Raspberry Pi 3 (Raspbian) @ 700MHz: 0.192s/frame </td>
+        <td colspan="3"> Raspberry Pi 3 (Raspbian) @ 700MHz : 171ms/frame </td>
     </tr>
     <tr>
         <td> Uncompressed Size </td>
@@ -41,7 +41,7 @@ permalink: /gallery/ILSVRC2012/Buckthorn
     </tr>
     <tr>
         <td> Notes </td>
-        <td colspan="3"> Trained by Juan Lema using PyTorch 0.4.0 </td>
+        <td colspan="3"> Trained by Byron Changuion using PyTorch 1.0.0 </td>
     </tr>
 </table>
 
diff --git a/docs/gallery/ILSVRC2012/Candlenut.md b/docs/gallery/ILSVRC2012/Candlenut.md
@@ -6,20 +6,20 @@ permalink: /gallery/ILSVRC2012/Candlenut
 
 [Back to Gallery](/ELL/gallery)
 
-## ILSVRC2012 Classification: 64x64x3 Convolutional Neural Network (54.39% top 1 accuracy, 78.02% top 5 accuracy, 0.166s/frame on Raspberry Pi 3 (Raspbian) @ 700MHz)
+## ILSVRC2012 Classification: 64x64x3 Convolutional Neural Network (54.39% top 1 accuracy, 78.02% top 5 accuracy, 146ms/frame on Raspberry Pi 3 (Raspbian) @ 700MHz)
 
 <table class="table table-striped table-bordered">
     <tr>
         <td> Download </td>
-        <td colspan="3"> <a href="https://github.com/Microsoft/ELL-models/raw/master/models/ILSVRC2012/Candlenut/Candlenut.ell.zip">Candlenut.ell.zip</a></td>
+        <td colspan="3"> <a href="https://github.com/Microsoft/ELL-models/raw/master/models/ILSVRC2012/Candlenut/Candlenut_pi3.ell.zip">Candlenut_pi3.ell.zip</a></td>
     </tr>
     <tr>
         <td> Accuracy </td>
         <td colspan="3"> ILSVRC2012: 78.02% (Top 5), 54.39% (Top 1) </td>
     </tr>
     <tr>
         <td> Performance </td>
-        <td colspan="3"> Raspberry Pi 3 (Raspbian) @ 700MHz: 0.166s/frame </td>
+        <td colspan="3"> Raspberry Pi 3 (Raspbian) @ 700MHz: 146ms/frame </td>
     </tr>
     <tr>
         <td> Uncompressed Size </td>
@@ -41,7 +41,7 @@ permalink: /gallery/ILSVRC2012/Candlenut
     </tr>
     <tr>
         <td> Notes </td>
-        <td colspan="3"> Trained by Juan Lema using CNTK 2.3.1 </td>
+        <td colspan="3"> Trained by Byron Changuion using PyTorch 1.0.0 </td>
     </tr>
 </table>
 
diff --git a/docs/gallery/ILSVRC2012/Carrot.md b/docs/gallery/ILSVRC2012/Carrot.md
@@ -6,20 +6,20 @@ permalink: /gallery/ILSVRC2012/Carrot
 
 [Back to Gallery](/ELL/gallery)
 
-## ILSVRC2012 Classification: 128x128x3 Convolutional Neural Network (64.61% top 1 accuracy, 85.63% top 5 accuracy, 0.444s/frame on Raspberry Pi 3 (Raspbian) @ 700MHz)
+## ILSVRC2012 Classification: 128x128x3 Convolutional Neural Network (64.61% top 1 accuracy, 85.63% top 5 accuracy, 397ms/frame on Raspberry Pi 3 (Raspbian) @ 700MHz)
 
 <table class="table table-striped table-bordered">
     <tr>
         <td> Download </td>
-        <td colspan="3"> <a href="https://github.com/Microsoft/ELL-models/raw/master/models/ILSVRC2012/Carrot/Carrot.ell.zip">Carrot.ell.zip</a></td>
+        <td colspan="3"> <a href="https://github.com/Microsoft/ELL-models/raw/master/models/ILSVRC2012/Carrot/Carrot_pi3.ell.zip">Carrot_pi3.ell.zip</a></td>
     </tr>
     <tr>
         <td> Accuracy </td>
         <td colspan="3"> ILSVRC2012: 85.63% (Top 5), 64.61% (Top 1) </td>
     </tr>
     <tr>
         <td> Performance </td>
-        <td colspan="3"> Raspberry Pi 3 (Raspbian) @ 700MHz: 0.444s/frame </td>
+        <td colspan="3"> Raspberry Pi 3 (Raspbian) @ 700MHz: 397ms/frame </td>
     </tr>
     <tr>
         <td> Uncompressed Size </td>
@@ -41,7 +41,7 @@ permalink: /gallery/ILSVRC2012/Carrot
     </tr>
     <tr>
         <td> Notes </td>
-        <td colspan="3"> Trained by Juan Lema using CNTK 2.3.1 </td>
+        <td colspan="3"> Trained by Byron Changuion using PyTorch 1.0.0 </td>
     </tr>
 </table>
 
diff --git a/docs/gallery/ILSVRC2012/CashewNut.md b/docs/gallery/ILSVRC2012/CashewNut.md
@@ -6,20 +6,20 @@ permalink: /gallery/ILSVRC2012/CashewNut
 
 [Back to Gallery](/ELL/gallery)
 
-## ILSVRC2012 Classification: 128x128x3 Convolutional Neural Network (60.22% top 1 accuracy, 82.44% top 5 accuracy, 0.176s/frame on Raspberry Pi 3 (Raspbian) @ 700MHz)
+## ILSVRC2012 Classification: 128x128x3 Convolutional Neural Network (60.22% top 1 accuracy, 82.44% top 5 accuracy, 178ms/frame on Raspberry Pi 3 (Raspbian) @ 700MHz)
 
 <table class="table table-striped table-bordered">
     <tr>
         <td> Download </td>
-        <td colspan="3"> <a href="https://github.com/Microsoft/ELL-models/raw/master/models/ILSVRC2012/CashewNut/CashewNut.ell.zip">CashewNut.ell.zip</a></td>
+        <td colspan="3"> <a href="https://github.com/Microsoft/ELL-models/raw/master/models/ILSVRC2012/CashewNut/CashewNut_pi3.ell.zip">CashewNut_pi3.ell.zip</a></td>
     </tr>
     <tr>
         <td> Accuracy </td>
         <td colspan="3"> ILSVRC2012: 82.44% (Top 5), 60.22% (Top 1) </td>
     </tr>
     <tr>
         <td> Performance </td>
-        <td colspan="3"> Raspberry Pi 3 (Raspbian) @ 700MHz: 0.176s/frame </td>
+        <td colspan="3"> Raspberry Pi 3 (Raspbian) @ 700MHz: 178ms/frame </td>
     </tr>
     <tr>
         <td> Uncompressed Size </td>
@@ -41,7 +41,7 @@ permalink: /gallery/ILSVRC2012/CashewNut
     </tr>
     <tr>
         <td> Notes </td>
-        <td colspan="3"> Trained by Juan Lema using PyTorch 0.4.0 </td>
+        <td colspan="3"> Trained by Byron Changuion using PyTorch 1.0.0 </td>
     </tr>
 </table>
 
diff --git a/docs/gallery/ILSVRC2012/Chalta.md b/docs/gallery/ILSVRC2012/Chalta.md
@@ -6,20 +6,20 @@ permalink: /gallery/ILSVRC2012/Chalta
 
 [Back to Gallery](/ELL/gallery)
 
-## ILSVRC2012 Classification: 64x64x3 Convolutional Neural Network (58.74% top 1 accuracy, 81.59% top 5 accuracy, 0.219s/frame on Raspberry Pi 3 (Raspbian) @ 700MHz)
+## ILSVRC2012 Classification: 64x64x3 Convolutional Neural Network (58.74% top 1 accuracy, 81.59% top 5 accuracy, 221ms/frame on Raspberry Pi 3 (Raspbian) @ 700MHz)
 
 <table class="table table-striped table-bordered">
     <tr>
         <td> Download </td>
-        <td colspan="3"> <a href="https://github.com/Microsoft/ELL-models/raw/master/models/ILSVRC2012/Chalta/Chalta.ell.zip">Chalta.ell.zip</a></td>
+        <td colspan="3"> <a href="https://github.com/Microsoft/ELL-models/raw/master/models/ILSVRC2012/Chalta/Chalta_pi3.ell.zip">Chalta_pi3.ell.zip</a></td>
     </tr>
     <tr>
         <td> Accuracy </td>
         <td colspan="3"> ILSVRC2012: 81.59% (Top 5), 58.74% (Top 1) </td>
     </tr>
     <tr>
         <td> Performance </td>
-        <td colspan="3"> Raspberry Pi 3 (Raspbian) @ 700MHz: 0.219s/frame </td>
+        <td colspan="3"> Raspberry Pi 3 (Raspbian) @ 700MHz: 221ms/frame </td>
     </tr>
     <tr>
         <td> Uncompressed Size </td>
@@ -41,7 +41,7 @@ permalink: /gallery/ILSVRC2012/Chalta
     </tr>
     <tr>
         <td> Notes </td>
-        <td colspan="3"> Trained by Juan Lema using CNTK 2.3.1 </td>
+        <td colspan="3"> Trained by Byron Changuion using PyTorch 1.0.0 </td>
     </tr>
 </table>
 
diff --git a/docs/gallery/ILSVRC2012/Chenet.md b/docs/gallery/ILSVRC2012/Chenet.md
@@ -6,7 +6,7 @@ permalink: /gallery/ILSVRC2012/Chenet
 
 [Back to Gallery](/ELL/gallery)
 
-## ILSVRC2012 Classification: 128x128x3 Convolutional Neural Network (68.19% top 1 accuracy, 88.24% top 5 accuracy, 0.85s/frame on Raspberry Pi 3)
+## ILSVRC2012 Classification: 128x128x3 Convolutional Neural Network (68.19% top 1 accuracy, 88.24% top 5 accuracy, 850ms/frame on Raspberry Pi 3)
 
 <table class="table table-striped table-bordered">
     <tr>
@@ -19,7 +19,7 @@ permalink: /gallery/ILSVRC2012/Chenet
     </tr>
     <tr>
         <td> Performance </td>
-        <td colspan="3"> Raspberry Pi 3 (Raspbian) @ 700MHz: 0.85s/frame </td>
+        <td colspan="3"> Raspberry Pi 3 (Raspbian) @ 700MHz: 850ms/frame </td>
     </tr>
     <tr>
         <td> Uncompressed Size </td>

diff --git a/docs/gallery/ILSVRC2012/Clary.md b/docs/gallery/ILSVRC2012/Clary.md
@@ -6,20 +6,20 @@ permalink: /gallery/ILSVRC2012/Clary
 
 [Back to Gallery](/ELL/gallery)
 
-## ILSVRC2012 Classification: 128x128x3 Convolutional Neural Network (66.65% top 1 accuracy, 87.17% top 5 accuracy, 0.597s/frame on Raspberry Pi 3 (Raspbian) @ 700MHz)
+## ILSVRC2012 Classification: 128x128x3 Convolutional Neural Network (66.65% top 1 accuracy, 87.17% top 5 accuracy, 506ms/frame on Raspberry Pi 3 (Raspbian) @ 700MHz)
 
 <table class="table table-striped table-bordered">
     <tr>
         <td> Download </td>
-        <td colspan="3"> <a href="https://github.com/Microsoft/ELL-models/raw/master/models/ILSVRC2012/Clary/Clary.ell.zip">Clary.ell.zip</a></td>
+        <td colspan="3"> <a href="https://github.com/Microsoft/ELL-models/raw/master/models/ILSVRC2012/Clary/Clary_pi3.ell.zip">Clary_pi3.ell.zip</a></td>
     </tr>
     <tr>
         <td> Accuracy </td>
         <td colspan="3"> ILSVRC2012: 87.17% (Top 5), 66.65% (Top 1) </td>
     </tr>
     <tr>
         <td> Performance </td>
-        <td colspan="3"> Raspberry Pi 3 (Raspbian) @ 700MHz: 0.597s/frame </td>
+        <td colspan="3"> Raspberry Pi 3 (Raspbian) @ 700MHz: 506ms/frame </td>
     </tr>
     <tr>
         <td> Uncompressed Size </td>
@@ -41,7 +41,7 @@ permalink: /gallery/ILSVRC2012/Clary
     </tr>
     <tr>
         <td> Notes </td>
-        <td colspan="3"> Trained by Juan Lema using PyTorch 0.4.0 </td>
+        <td colspan="3"> Trained by Byron Changuion using PyTorch 1.0.0 </td>
     </tr>
 </table>
 
diff --git a/docs/gallery/ILSVRC2012/Clover.md b/docs/gallery/ILSVRC2012/Clover.md
@@ -6,20 +6,20 @@ permalink: /gallery/ILSVRC2012/Clover
 
 [Back to Gallery](/ELL/gallery)
 
-## ILSVRC2012 Classification: 64x64x3 Convolutional Neural Network (53.04% top 1 accuracy, 77.12% top 5 accuracy, 0.145s/frame on Raspberry Pi 3 (Raspbian) @ 700MHz)
+## ILSVRC2012 Classification: 64x64x3 Convolutional Neural Network (53.04% top 1 accuracy, 77.12% top 5 accuracy, 126ms/frame on Raspberry Pi 3 (Raspbian) @ 700MHz)
 
 <table class="table table-striped table-bordered">
     <tr>
         <td> Download </td>
-        <td colspan="3"> <a href="https://github.com/Microsoft/ELL-models/raw/master/models/ILSVRC2012/Clover/Clover.ell.zip">Clover.ell.zip</a></td>
+        <td colspan="3"> <a href="https://github.com/Microsoft/ELL-models/raw/master/models/ILSVRC2012/Clover/Clover_pi3.ell.zip">Clover_pi3.ell.zip</a></td>
     </tr>
     <tr>
         <td> Accuracy </td>
         <td colspan="3"> ILSVRC2012: 77.12% (Top 5), 53.04% (Top 1) </td>
     </tr>
     <tr>
         <td> Performance </td>
-        <td colspan="3"> Raspberry Pi 3 (Raspbian) @ 700MHz: 0.145s/frame </td>
+        <td colspan="3"> Raspberry Pi 3 (Raspbian) @ 700MHz: 126ms/frame </td>
     </tr>
     <tr>
         <td> Uncompressed Size </td>
@@ -41,7 +41,7 @@ permalink: /gallery/ILSVRC2012/Clover
     </tr>
     <tr>
         <td> Notes </td>
-        <td colspan="3"> Trained by Juan Lema using PyTorch 0.4.0 </td>
+        <td colspan="3"> Trained by Byron Changuion using PyTorch 1.0.0 </td>
     </tr>
 </table>
 
diff --git a/docs/gallery/ILSVRC2012/Coconut.md b/docs/gallery/ILSVRC2012/Coconut.md
@@ -6,20 +6,20 @@ permalink: /gallery/ILSVRC2012/Coconut
 
 [Back to Gallery](/ELL/gallery)
 
-## ILSVRC2012 Classification: 64x64x3 Convolutional Neural Network (44.41% top 1 accuracy, 69.41% top 5 accuracy, 0.049s/frame on Raspberry Pi 3 (Raspbian) @ 700MHz)
+## ILSVRC2012 Classification: 64x64x3 Convolutional Neural Network (44.41% top 1 accuracy, 69.41% top 5 accuracy, 46ms/frame on Raspberry Pi 3 (Raspbian) @ 700MHz)
 
 <table class="table table-striped table-bordered">
     <tr>
-        <td> Download </td>
-        <td colspan="3"> <a href="https://github.com/Microsoft/ELL-models/raw/master/models/ILSVRC2012/Coconut/Coconut.ell.zip">Coconut.ell.zip</a></td>
+        <td> Download  </td>
+        <td colspan="3"> <a href="https://github.com/Microsoft/ELL-models/raw/master/models/ILSVRC2012/Coconut/Coconut_pi3.ell.zip">Coconut_pi3.ell.zip</a></td>
     </tr>
     <tr>
         <td> Accuracy </td>
         <td colspan="3"> ILSVRC2012: 69.41% (Top 5), 44.41% (Top 1) </td>
     </tr>
     <tr>
         <td> Performance </td>
-        <td colspan="3"> Raspberry Pi 3 (Raspbian) @ 700MHz: 0.049s/frame </td>
+        <td colspan="3"> Raspberry Pi 3 (Raspbian) @ 700MHz: 46ms/frame </td>
     </tr>
     <tr>
         <td> Uncompressed Size </td>
@@ -41,7 +41,7 @@ permalink: /gallery/ILSVRC2012/Coconut
     </tr>
     <tr>
         <td> Notes </td>
-        <td colspan="3"> Trained by Juan Lema using PyTorch 0.4.0 </td>
+        <td colspan="3"> Trained by Byron Changuion using PyTorch 1.0.0 </td>
     </tr>
 </table>
 
diff --git a/docs/gallery/ILSVRC2012/Elderflower.md b/docs/gallery/ILSVRC2012/Elderflower.md
@@ -6,7 +6,7 @@ permalink: /gallery/ILSVRC2012/Elderflower
 
 [Back to Gallery](/ELL/gallery)
 
-## ILSVRC2012 Classification: 64x64x3 Convolutional Neural Network (49.32% top 1 accuracy, 73.99% top 5 accuracy, 0.109s/frame on Raspberry Pi 3 (Raspbian) @ 700MHz)
+## ILSVRC2012 Classification: 64x64x3 Convolutional Neural Network (49.32% top 1 accuracy, 73.99% top 5 accuracy, 112ms/frame on Raspberry Pi 3 (Raspbian) @ 700MHz)
 
 <table class="table table-striped table-bordered">
     <tr>
@@ -19,7 +19,7 @@ permalink: /gallery/ILSVRC2012/Elderflower
     </tr>
     <tr>
         <td> Performance </td>
-        <td colspan="3"> Raspberry Pi 3 (Raspbian) @ 700MHz: 0.109s/frame </td>
+        <td colspan="3"> Raspberry Pi 3 (Raspbian) @ 700MHz: 112ms/frame </td>
     </tr>
     <tr>
         <td> Uncompressed Size </td>

diff --git a/docs/gallery/ILSVRC2012/GardenCress.md b/docs/gallery/ILSVRC2012/GardenCress.md
@@ -6,7 +6,7 @@ permalink: /gallery/ILSVRC2012/GardenCress
 
 [Back to Gallery](/ELL/gallery)
 
-## ILSVRC2012 Classification: 128x128x3 Convolutional Neural Network (55.00% top 1 accuracy, 78.75% top 5 accuracy, 0.17s/frame on Raspberry Pi 3)
+## ILSVRC2012 Classification: 128x128x3 Convolutional Neural Network (55.00% top 1 accuracy, 78.75% top 5 accuracy, 170ms/frame on Raspberry Pi 3)
 
 <table class="table table-striped table-bordered">
     <tr>
@@ -19,7 +19,7 @@ permalink: /gallery/ILSVRC2012/GardenCress
     </tr>
     <tr>
         <td> Performance </td>
-        <td colspan="3"> Raspberry Pi 3 (Raspbian) @ 700MHz: 0.17s/frame </td>
+        <td colspan="3"> Raspberry Pi 3 (Raspbian) @ 700MHz: 170ms/frame </td>
     </tr>
     <tr>
         <td> Uncompressed Size </td>