This paper assesses the performance of using pivoted datasets for Bangla text summarizing as compared to human-generated datasets by a fine-tuned transformer model such as mT5-small, BanglaT5 or mBART. The pivoted dataset is created by translating Bangla text to English, summarizing it using a pre-trained T5-small model and then translating back to Bangla. The goal is to assess whether the pivoted data can effectively substitute human-generated data for a low-resource language like Bangla. The XLsum dataset that contains Bangla and English article-summary pairs from BBC is used to create synthetic datasets through the pivoting technique. The models are fine-tuned on both original and synthetic datasets by maintaining consistent training parameters. Summarization performance is evaluated using the BLEU, METEOR, and chrF++ scores. Experimental results reveal that pivoted datasets achieve around 85% of the performance of human-generated datasets on average in METEOR and chrF++ scores. In terms of BLEU scores, the performance is comparable when human generated data is combined with the pivoted data in different models. Our findings suggest that while human-generated data provides a better performance, the pivoted datasets can be viably used for summarizing tasks in a low-resource language, like Bangla.