Add prelim results draft

docs/index.md

@@ -5,11 +5,43 @@ title: Home
 
 # Flow on the Go
 
-<br>
+### {{ site.description }}
 
-<a class="btn theme-base-0b" href="{{ site.baseurl }}/proposal">Proposal</a>
+Preliminary Results (Updated May 10)
 
-<br>
+[Ashwin Sekar](mailto:[email protected]) (asekar)
+and [Richard Zhao](mailto:[email protected]) (richardz)
 
-<a class="btn theme-base-0b" href="{{ site.baseurl }}/checkpoint">Checkpoint</a>
+![frames]({{ site.baseurl }}/public/img/temple_3.gif)
+![flow]({{ site.baseurl }}/public/img/flow.gif)
 
+## Summary
+
+We implement super-realtime (>30fps), high resolution optical flows on a mobile GPU platform. Fast
+optical flows allow a realtime video processing pipeline to use the flow in other algorithms such
+as object detection or image stabilization.
+
+## Challenges
+
+Image pyramids
+
+Number of patches increases
+
+Maintaining accuracy of the flow
+
+Memory management
+
+Optimizing to Jetson (which has a lackluster CPU)
+
+## Preliminary Results
+
+All results are from our code running on an NVIDIA Jetson TX2.
+
+### Optical Flow (total)
+
+Using a hybrid GPU-CPU implementation, we achieve an end-to-end latency of roughly 10ms. This
+is a speedup of roughly 10x.
+
+### Image pyramid construction
+
+90 ms => 3 ms (30x speedup)