Skip to main content
. Author manuscript; available in PMC: 2020 Sep 28.
Published in final edited form as: Conf Comput Vis Pattern Recognit Workshops. 2020 Jul 28;2020:3639–3648. doi: 10.1109/cvprw50498.2020.00425

Table 4.

Comparison of the average time taken to compute PIs for one image using conventional TDA tools and the proposed PI-Net model. The time reported is averaged over all images present in. the training set of each dataset.

Method Time (10−3 seconds)

CIFAR10 (50000 images) SVHN (73257 images)
Conventional TDA - CPU 3567.29±867.74 3433.06±622.21
PI-Net - CPU 125.45±5.30 125.49±5.34
PI-Net - GPU 2.52±0.02 2.19±0.02