Objectives: Employing computer vision methods for yield estimation requires standard datasets. However, unavailability of such a dataset in the literature limits the researchers from evaluating and comparing the performance of their proposed algorithms. In this paper, a benchmark of mango tree dataset is introduced. Methods: The dataset is gathered over a 5-month period, starting from the fruit's blossoming stage and ending with its ripening stage. There are 21,000 photos of 4 distinct mango tree varieties in the dataset. From each cultivar, images are captured with different views, distances, and daylight conditions. Further, preprocessing, and exploratory analysis of the dataset are carried out by extracting a few global features such as colors, textures, and histograms for intra-class and inter-class mango trees. Findings: The analysis of the collected dataset with different color layers by extracting a few global features and classification of the cultivars of mango trees and, based on the results obtained, the optimal layer of color is attained for the further yield estimation process. Novelty: Exploratory analysis of a novel temporal mango crop dataset is executed, and a color analysis classification method is proposed to aid in the early estimation of mango fruit crop yields. Keywords: Mango Fruit, Yield Estimation, Temporal Dataset, Feature Extraction, Tree Classification