(MB) 93 320 32 / 50Hand gesture recognition 33 / 50Data ◮ Google MediaPipe’s Rock-Paper-Scissors dataset for hand gesture recognition ◮ Contains 125 images for each class ◮ Images have various sizes ◮ ◮ Images have 3 color channels per pixel (RGB) ◮ Laurence Moroney’s Rock-Paper-Scissors dataset (published by Sani Kamal on Kaggle) ◮ Contains 964 images for each class split into training data (840) in Python, on a laptop Dataset Train duration Model size (MB) Small 2m15s 23 Big 7m10s 23 36 / 50Test the Rock-Paper-Scissors model ◮ Using the testing data of the big dataset (124 images per class)

5How DoD is used in actual code • Platform specific code • Procedural/imperative code • Problem/dataset specific code • Hand-optimize cache lines and struct layout 6How can we improve it? • We could opposite philosophy Data oriented code • Platform specific code • Procedural/imperative code • Problem/dataset specific code • Hand-optimize cache lines and struct layout 9 Opposite philosophy • Cross platform opposite philosophy Data oriented code • Platform specific code • Procedural/imperative code • Problem/dataset specific code • Hand-optimize cache lines and struct layout 10 Opposite philosophy • Cross

are all of them sorted by how often they appear as user-defined identifiers in that particular dataset: …” assertexpr 0 ccassert 0 co_assert 0 contract_assert 0 contractassert 0 cppassert are all of them sorted by how often they appear as user-defined identifiers in that particular dataset: …” assertexpr 0 ccassert 0 co_assert 0 contract_assert 0 contractassert 0 cppassert 3734 matches … but we also did *not* make yield a full keyword, which has 9533 matches in that dataset, and chose co_yield instead… Where is the bar?”76 (1) Naming is hard (2) Compatibility is a major