[Weekly Review] 2021/12/13-2021/12/19
2021/12/13-2021/12/19
Last week, I mainly focused on the performance testing of ShapoolNMS using Chisel implementations. I fixed one bug. I also tried to analysis the score distribution after log2
. My cooperates from ASTAR said maybe we can try to output float-point scores from softmax
instead of fixed-point for better theoretically higher mAP.
I continued learn formal and architecture on Saturday.