Algorithm 1: Proposed computational TTA approach on the test sample images.

Input: The Pre-trained Model is defined as ‘M’ on in-distribution data. The augmented terminology is denoted as ‘A’. The test samples set of images data $X_1, X_{2 }{,X}_3, \dots {,X}_n$; which contains in-distribution data and out-distribution data combinedly. Output: Predicted samples with distributed mean score as              $S{(X}_1$), $S{(X}_2$), $S{(X}_3$), …, $S{(X}_n$); 1: Begin 2: For loop is executed: i = 1 to n do 3:    Data augmentation for each input sample set, $X_i$= $\{X_1, X_{2 }{,X}_3, \dots {,X}_n$}; 4:    Converted to horizontal flip (HF) with M to obtain ${A(X}_i),$ respectively. Similarly, vertical flip (VF) is generated for each sample to ${A(X}_i)$; 5:    Feed $X_i$ to ${A(X}_i)$ into model M, continuously calculate ${M(X}_i), {M\left(A\right(X}_i))$ through forward passes; 6:    Mean result, $S(X_i)=\frac{\sum _{i=1}^n{M\left(A\right(X}_i\left)\right) }{n}$; 7: end for 8: End