{\rtf1\ansi\ansicpg1252\cocoartf1671\cocoasubrtf600 {\fonttbl\f0\fswiss\fcharset0 Helvetica;} {\colortbl;\red255\green255\blue255;} {\*\expandedcolortbl;;} \margl1440\margr1440\vieww10800\viewh8400\viewkind0 \pard\tx720\tx1440\tx2160\tx2880\tx3600\tx4320\tx5040\tx5760\tx6480\tx7200\tx7920\tx8640\pardirnatural\partightenfactor0 \f0\fs24 \cf0
Part 1
\Here are some sample images from my dataloader with ground truth keypoints:
\ \Train (red) and val (blue) loss with learning rate 0.001
\ \Train (red) and val (blue) loss with learning rate 0.0001
\ \Examples where the nose was detected correctly:
\ \Examples where the nose was detected incorrectly:
\ \In both of these examples it seems as though other artifacts on the subjects' faces are being recognized as a nose.
\Part 2
\Part 3
\ }