Table 3.

Model evaluation with previous work including the used approach, optimizer, learning rate, batch size, loss function, and accuracy.

Ref	Approach	Optimizer	Learning Rate	Batch Size	Loss Function	Accuracy
Chan et al. [27]	Geometry-aware 3D using GAN	Adam	0.0002	32	Cross-entropy	87.2%
Xing et al. [23]	Deep Q-learning approximates the Q-value function, which estimates the expected cumulative reward for taking a particular action in a particular state	Adam	0.001	32	Q-learning with mean squared error (MSE)	Average reward
Hetzel et al. [24]	RL with DDPG for virtual keyboard typing	Adam	0.001	64	DDPG	Average reward
Our Model	RL-PipTrack: AR-assisted deep RL-based model	PPO	0.0005	10	Clipped surrogate objective	Average reward + standard deviation