Described results of Bayesian SSD without dropout

Signed-off-by: Jim Martens <github@2martens.de>
This commit is contained in:
Jim Martens 2019-09-09 12:09:05 +02:00
parent 9fde66abf1
commit 0ac3176626
1 changed files with 12 additions and 1 deletions

View File

@ -652,7 +652,9 @@ Bayesian SSD was run with 0.2 confidence threshold and compared
to vanilla SSD with 0.2 confidence threshold. Coupled with the
entropy threshold, this comparison shows how uncertain the network
is. If it is very certain the dropout sampling should have no
significant impact on the result.
significant impact on the result. Furthermore, in two cases the
dropout was turned off to isolate the impact of non-maximum suppression
on the result.
Both, vanilla SSD with entropy thresholding and Bayesian SSD with
entropy thresholding, were tested for entropy thresholds ranging
@ -708,6 +710,15 @@ not very uncertain. The best performing entropy threshold is not any better than
the corresponding vanilla SSD without entropy threshold. Therefore, in this
case the per-class confidence score is far more important for the result.
The results for Bayesian SSD show a massive impact of the existance of
non-maximum suppression: maximum \(F_1\) score of 0.371 (with NMS) to 0.006 (without NMS)
with micro averaging and 0.363 (with NMS) to 0.006 (without NMS) with macro averaging.
Dropout was disabled in both cases, making them effectively a vanilla SSD run
with multiple forward passes. Therefore, the low number of open set errors with
micro averaging (164 without NMS) does not qualify as a good result and is not
marked bold, although it is the lowest number.
\begin{table}[ht]
\begin{tabular}{rcccc}
\hline