Table 4 Comparison of P2M with related in-sensor and near-sensor computing works.

From: A processing-in-pixel-in-memory paradigm for resource-constrained TinyML applications

Work

Tech node

Computation

High resolution

Dataset

Supported Ops.

Acc.(\(\%\))

P2M (ours)

22 nm

Analog

Yes

VWW

Conv, BN, ReLU

89.90

TCAS-I 202010

180 nm

Analog

No

–

Binary Conv.

–

TCSVT 202213

180 nm

Analog

No

CIFAR-10

Conv.

89.6

Nature 202011

–

Analog

No

3-class alphabet

MLP

100

ECCV 202012

180 nm

Digital

No

MNIST

MLP

93.0

Back to article page