Faster GPU NMS operator (#16542)

* Adding second NMS op * NMS kernel * Removing second sort * Optimization * Adding out-of-place ability to SortByKey * Optimization pt2 * Optimizations pt3 * Do not recompute other boxes area every time * Sort only topk results during second sorting * Cleaning * Fixes from rebase * Fix lint and more fixes from rebase * Fix typo * Early exit in Triangle kernel * Fixes * Fix sort * Fix from rebase * Fix for the mixed naming convention * Fix the index_t with int comparisoon
apache · Nov 5, 2019 · 0c5677e · 0c5677e
1 parent c583e44
commit 0c5677e
Show file tree

Hide file tree

Showing 4 changed files with 840 additions and 38 deletions.
diff --git a/src/operator/contrib/bounding_box.cc b/src/operator/contrib/bounding_box.cc
@@ -34,6 +34,7 @@ DMLC_REGISTER_PARAMETER(BoxOverlapParam);
 DMLC_REGISTER_PARAMETER(BipartiteMatchingParam);
 DMLC_REGISTER_PARAMETER(BoxDecodeParam);
 
+
 NNVM_REGISTER_OP(_contrib_box_nms)
 .add_alias("_contrib_box_non_maximum_suppression")
 .describe(R"code(Apply non-maximum suppression to input.