Robustness and Explainability of Image Classification Based on QCNN

Table 1

Experimental results of different bitplanes on MERA and boxcounting features.

Feature	MERA accuracy	Box accuracy	MERA loss	Box loss
Original	0.9531	0.9844	0.2526	0.0837
Bitplane1	0.7246	0.7326	0.2353	0.3046
Bitplane2	0.7244	0.7355	0.2839	0.2101
Bitplane3	0.7203	0.7364	0.2892	0.2299
Bitplane4	0.7348	0.7552	0.2486	0.2869
Bitplane5	0.9688	0.8906	0.2118	0.7642
Bitplane6	0.9844	0.9062	0.2044	0.2981
Bitplane7	0.9844	0.8281	0.6144	0.7873
Bitplane8	0.9531	0.9821	0.7535	0.6259