Extended Data Fig. 3: Sensitivity analysis of our method. | Nature

Extended Data Fig. 3: Sensitivity analysis of our method.

From: Low-latency automotive vision with event cameras

Extended Data Fig. 3

(a) Performance of DAGr+ResNet-50 on the DSEC-Detection test-set for different event-to-image synchronization errors. (b) Object detection score of our method for differently interpolated ground truth. Here we use the DSEC-Detection subset with object tracks with a length of at least four. (c) Performance of different CNN backbones with DAGr-S on the full DSEC-Detection test set, with and without CNN latency considered. All performances are measured in mAP (higher is better). Speed refers to the average runtime of the CNN alone over images from the test set.

Back to article page