Add calib uncertainties in object localization

Author: patrick-llgc

paper_notes/2dod_calib.md

@@ -0,0 +1,19 @@
+# [Calibrating Uncertainties in Object Localization Task](https://arxiv.org/abs/1811.11210)
+
+_November 2019_
+
+tl;dr: Proof of concept by applying Uncertainty calibration to object detector. 
+
+#### Overall impression
+For a more theoretical treatment refer to [accurate uncertainty via calibrated regression](dl_regression_calib.md). A more detailed application is [can we trust you](towards_safe_ad_calib.md).
+
+#### Key ideas
+- Validating uncertainty estimates: plot regressed aleatoric uncertainty $\sigma_i^2$ and $(b_i - \bar{b_i})^2$
+- To find 90% confidence interval, the upper and lower bounds are given by $\hat{P^{-1}}(r \pm 90/2)$, where $r = \hat{P(x)}$ and $\hat{P}$ is the P after calibration.
+
+#### Technical details
+- Summary of technical details
+
+#### Notes
+- Questions and notes on how to improve/revise the current work  
+

paper_notes/towards_safe_ad_calib.md

@@ -7,7 +7,7 @@ tl;dr: Calibration of the network for a probabilistic object detector
 #### Overall impression
 The paper extends previous works in the [probabilistic lidar detector](towards_safe_ad.md) and its [successor](towards_safe_ad2.md). It is based on the work of Pixor. 
 
-Calibration: a probabilistic object detector should predict uncertainties that match the natural frequency of correct predictions. 90% of the predictions with 0.9 score from a calibrated detector should be correct. Humans have intuitive notion of probability in a frequentist sense. --> cf [accurate uncertainty via calibrated regression](dl_regression_calib.md).
+Calibration: a probabilistic object detector should predict uncertainties that match the natural frequency of correct predictions. 90% of the predictions with 0.9 score from a calibrated detector should be correct. Humans have intuitive notion of probability in a frequentist sense. --> cf [accurate uncertainty via calibrated regression](dl_regression_calib.md) and [calib uncertainties in object detection](2dod_calib.md).
 
 A calibrated regression is a bit harder to interpret. P(gt < F^{-1}(p)) = p. F^{-1} = F_q is the inverse function of CDF, the quantile function.